commit: use `dirstate.change_files` to scope the associated `addremove`

This was significantly more complicated than I expected, because multiple
extensions get in the way.

I introduced a context that lazily open the transaction and associated context
to work around these complication. See the inline documentation for details.

Introducing the wrapping transaction remove the need for dirstate-guard (one of
the ultimate goal of all this), and slightly affect the result of a `hg
rollback` after a `hg commit --addremove`. That last part is deemed fine. It
aligns the behavior with what happens after a failed `hg commit --addremove` and
nobody should be using `hg rollback` anyway.

The small output change in the test come from the different transaction timing
and fact the transaction now backup the dirstate before the addremove, which
might mean "no file to backup" when the repository starts from an empty state.

hgext/largefiles/overrides.py

@@ -1,1890 +1,1903 b''
 # Copyright 2009-2010 Gregory P. Ward
 # Copyright 2009-2010 Intelerad Medical Systems Incorporated
 # Copyright 2010-2011 Fog Creek Software
 # Copyright 2010-2011 Unity Technologies
 #
 # This software may be used and distributed according to the terms of the
 # GNU General Public License version 2 or any later version.
 
 '''Overridden Mercurial commands and functions for the largefiles extension'''
 
 import contextlib
 import copy
 import os
 
 from mercurial.i18n import _
 
 from mercurial.pycompat import open
 
 from mercurial.hgweb import webcommands
 
 from mercurial import (
     archival,
     cmdutil,
     copies as copiesmod,
     dirstate,
     error,
     exchange,
     extensions,
     exthelper,
     filemerge,
     hg,
     logcmdutil,
     match as matchmod,
     merge,
     mergestate as mergestatemod,
     pathutil,
     pycompat,
     scmutil,
     smartset,
     subrepo,
     url as urlmod,
     util,
 )
 
 from mercurial.upgrade_utils import (
     actions as upgrade_actions,
 )
 
 from . import (
     lfcommands,
     lfutil,
     storefactory,
 )
 
 ACTION_ADD = mergestatemod.ACTION_ADD
 ACTION_DELETED_CHANGED = mergestatemod.ACTION_DELETED_CHANGED
 ACTION_GET = mergestatemod.ACTION_GET
 ACTION_KEEP = mergestatemod.ACTION_KEEP
 ACTION_REMOVE = mergestatemod.ACTION_REMOVE
 
 eh = exthelper.exthelper()
 
 lfstatus = lfutil.lfstatus
 
 MERGE_ACTION_LARGEFILE_MARK_REMOVED = mergestatemod.MergeAction('lfmr')
 
 # -- Utility functions: commonly/repeatedly needed functionality ---------------
 
 
 def composelargefilematcher(match, manifest):
     """create a matcher that matches only the largefiles in the original
     matcher"""
     m = copy.copy(match)
     lfile = lambda f: lfutil.standin(f) in manifest
     m._files = [lf for lf in m._files if lfile(lf)]
     m._fileset = set(m._files)
     m.always = lambda: False
     origmatchfn = m.matchfn
     m.matchfn = lambda f: lfile(f) and origmatchfn(f)
     return m
 
 
 def composenormalfilematcher(match, manifest, exclude=None):
     excluded = set()
     if exclude is not None:
         excluded.update(exclude)
 
     m = copy.copy(match)
     notlfile = lambda f: not (
         lfutil.isstandin(f) or lfutil.standin(f) in manifest or f in excluded
     )
     m._files = [lf for lf in m._files if notlfile(lf)]
     m._fileset = set(m._files)
     m.always = lambda: False
     origmatchfn = m.matchfn
     m.matchfn = lambda f: notlfile(f) and origmatchfn(f)
     return m
 
 
 def addlargefiles(ui, repo, isaddremove, matcher, uipathfn, **opts):
     large = opts.get('large')
     lfsize = lfutil.getminsize(
         ui, lfutil.islfilesrepo(repo), opts.get('lfsize')
     )
 
     lfmatcher = None
     if lfutil.islfilesrepo(repo):
         lfpats = ui.configlist(lfutil.longname, b'patterns')
         if lfpats:
             lfmatcher = matchmod.match(repo.root, b'', list(lfpats))
 
     lfnames = []
     m = matcher
 
     wctx = repo[None]
     for f in wctx.walk(matchmod.badmatch(m, lambda x, y: None)):
         exact = m.exact(f)
         lfile = lfutil.standin(f) in wctx
         nfile = f in wctx
         exists = lfile or nfile
 
         # Don't warn the user when they attempt to add a normal tracked file.
         # The normal add code will do that for us.
         if exact and exists:
             if lfile:
                 ui.warn(_(b'%s already a largefile\n') % uipathfn(f))
             continue
 
         if (exact or not exists) and not lfutil.isstandin(f):
             # In case the file was removed previously, but not committed
             # (issue3507)
             if not repo.wvfs.exists(f):
                 continue
 
             abovemin = (
                 lfsize and repo.wvfs.lstat(f).st_size >= lfsize * 1024 * 1024
             )
             if large or abovemin or (lfmatcher and lfmatcher(f)):
                 lfnames.append(f)
                 if ui.verbose or not exact:
                     ui.status(_(b'adding %s as a largefile\n') % uipathfn(f))
 
     bad = []
 
     # Need to lock, otherwise there could be a race condition between
     # when standins are created and added to the repo.
     with repo.wlock():
         if not opts.get('dry_run'):
             standins = []
             lfdirstate = lfutil.openlfdirstate(ui, repo)
             for f in lfnames:
                 standinname = lfutil.standin(f)
                 lfutil.writestandin(
                     repo,
                     standinname,
                     hash=b'',
                     executable=lfutil.getexecutable(repo.wjoin(f)),
                 )
                 standins.append(standinname)
                 lfdirstate.set_tracked(f)
             lfdirstate.write(repo.currenttransaction())
             bad += [
                 lfutil.splitstandin(f)
                 for f in repo[None].add(standins)
                 if f in m.files()
             ]
 
         added = [f for f in lfnames if f not in bad]
     return added, bad
 
 
 def removelargefiles(ui, repo, isaddremove, matcher, uipathfn, dryrun, **opts):
     after = opts.get('after')
     m = composelargefilematcher(matcher, repo[None].manifest())
     with lfstatus(repo):
         s = repo.status(match=m, clean=not isaddremove)
     manifest = repo[None].manifest()
     modified, added, deleted, clean = [
         [f for f in list if lfutil.standin(f) in manifest]
         for list in (s.modified, s.added, s.deleted, s.clean)
     ]
 
     def warn(files, msg):
         for f in files:
             ui.warn(msg % uipathfn(f))
         return int(len(files) > 0)
 
     if after:
         remove = deleted
         result = warn(
             modified + added + clean, _(b'not removing %s: file still exists\n')
         )
     else:
         remove = deleted + clean
         result = warn(
             modified,
             _(
                 b'not removing %s: file is modified (use -f'
                 b' to force removal)\n'
             ),
         )
         result = (
             warn(
                 added,
                 _(
                     b'not removing %s: file has been marked for add'
                     b' (use forget to undo)\n'
                 ),
             )
             or result
         )
 
     # Need to lock because standin files are deleted then removed from the
     # repository and we could race in-between.
     with repo.wlock():
         lfdirstate = lfutil.openlfdirstate(ui, repo)
         for f in sorted(remove):
             if ui.verbose or not m.exact(f):
                 ui.status(_(b'removing %s\n') % uipathfn(f))
 
             if not dryrun:
                 if not after:
                     repo.wvfs.unlinkpath(f, ignoremissing=True)
 
         if dryrun:
             return result
 
         remove = [lfutil.standin(f) for f in remove]
         # If this is being called by addremove, let the original addremove
         # function handle this.
         if not isaddremove:
             for f in remove:
                 repo.wvfs.unlinkpath(f, ignoremissing=True)
         repo[None].forget(remove)
 
         for f in remove:
             lfdirstate.set_untracked(lfutil.splitstandin(f))
 
         lfdirstate.write(repo.currenttransaction())
 
     return result
 
 
 # For overriding mercurial.hgweb.webcommands so that largefiles will
 # appear at their right place in the manifests.
 @eh.wrapfunction(webcommands, b'decodepath')
 def decodepath(orig, path):
     return lfutil.splitstandin(path) or path
 
 
 # -- Wrappers: modify existing commands --------------------------------
 
 
 @eh.wrapcommand(
     b'add',
     opts=[
         (b'', b'large', None, _(b'add as largefile')),
         (b'', b'normal', None, _(b'add as normal file')),
         (
             b'',
             b'lfsize',
             b'',
             _(
                 b'add all files above this size (in megabytes) '
                 b'as largefiles (default: 10)'
             ),
         ),
     ],
 )
 def overrideadd(orig, ui, repo, *pats, **opts):
     if opts.get('normal') and opts.get('large'):
         raise error.Abort(_(b'--normal cannot be used with --large'))
     return orig(ui, repo, *pats, **opts)
 
 
 @eh.wrapfunction(cmdutil, b'add')
 def cmdutiladd(orig, ui, repo, matcher, prefix, uipathfn, explicitonly, **opts):
     # The --normal flag short circuits this override
     if opts.get('normal'):
         return orig(ui, repo, matcher, prefix, uipathfn, explicitonly, **opts)
 
     ladded, lbad = addlargefiles(ui, repo, False, matcher, uipathfn, **opts)
     normalmatcher = composenormalfilematcher(
         matcher, repo[None].manifest(), ladded
     )
     bad = orig(ui, repo, normalmatcher, prefix, uipathfn, explicitonly, **opts)
 
     bad.extend(f for f in lbad)
     return bad
 
 
 @eh.wrapfunction(cmdutil, b'remove')
 def cmdutilremove(
     orig, ui, repo, matcher, prefix, uipathfn, after, force, subrepos, dryrun
 ):
     normalmatcher = composenormalfilematcher(matcher, repo[None].manifest())
     result = orig(
         ui,
         repo,
         normalmatcher,
         prefix,
         uipathfn,
         after,
         force,
         subrepos,
         dryrun,
     )
     return (
         removelargefiles(
             ui, repo, False, matcher, uipathfn, dryrun, after=after, force=force
         )
         or result
     )
 
 
 @eh.wrapfunction(dirstate.dirstate, b'_changing')
 @contextlib.contextmanager
 def _changing(orig, self, repo, change_type):
     pre = sub_dirstate = getattr(self, '_sub_dirstate', None)
     try:
         lfd = getattr(self, '_large_file_dirstate', False)
         if sub_dirstate is None and not lfd:
             sub_dirstate = lfutil.openlfdirstate(repo.ui, repo)
             self._sub_dirstate = sub_dirstate
         if not lfd:
             assert self._sub_dirstate is not None
         with orig(self, repo, change_type):
             if sub_dirstate is None:
                 yield
             else:
                 with sub_dirstate._changing(repo, change_type):
                     yield
     finally:
         self._sub_dirstate = pre
 
 
 @eh.wrapfunction(subrepo.hgsubrepo, b'status')
 def overridestatusfn(orig, repo, rev2, **opts):
     with lfstatus(repo._repo):
         return orig(repo, rev2, **opts)
 
 
 @eh.wrapcommand(b'status')
 def overridestatus(orig, ui, repo, *pats, **opts):
     with lfstatus(repo):
         return orig(ui, repo, *pats, **opts)
 
 
 @eh.wrapfunction(subrepo.hgsubrepo, b'dirty')
 def overridedirty(orig, repo, ignoreupdate=False, missing=False):
     with lfstatus(repo._repo):
         return orig(repo, ignoreupdate=ignoreupdate, missing=missing)
 
 
 @eh.wrapcommand(b'log')
 def overridelog(orig, ui, repo, *pats, **opts):
     def overridematchandpats(
         orig,
         ctx,
         pats=(),
         opts=None,
         globbed=False,
         default=b'relpath',
         badfn=None,
     ):
         """Matcher that merges root directory with .hglf, suitable for log.
         It is still possible to match .hglf directly.
         For any listed files run log on the standin too.
         matchfn tries both the given filename and with .hglf stripped.
         """
         if opts is None:
             opts = {}
         matchandpats = orig(ctx, pats, opts, globbed, default, badfn=badfn)
         m, p = copy.copy(matchandpats)
 
         if m.always():
             # We want to match everything anyway, so there's no benefit trying
             # to add standins.
             return matchandpats
 
         pats = set(p)
 
         def fixpats(pat, tostandin=lfutil.standin):
             if pat.startswith(b'set:'):
                 return pat
 
             kindpat = matchmod._patsplit(pat, None)
 
             if kindpat[0] is not None:
                 return kindpat[0] + b':' + tostandin(kindpat[1])
             return tostandin(kindpat[1])
 
         cwd = repo.getcwd()
         if cwd:
             hglf = lfutil.shortname
             back = util.pconvert(repo.pathto(hglf)[: -len(hglf)])
 
             def tostandin(f):
                 # The file may already be a standin, so truncate the back
                 # prefix and test before mangling it.  This avoids turning
                 # 'glob:../.hglf/foo*' into 'glob:../.hglf/../.hglf/foo*'.
                 if f.startswith(back) and lfutil.splitstandin(f[len(back) :]):
                     return f
 
                 # An absolute path is from outside the repo, so truncate the
                 # path to the root before building the standin.  Otherwise cwd
                 # is somewhere in the repo, relative to root, and needs to be
                 # prepended before building the standin.
                 if os.path.isabs(cwd):
                     f = f[len(back) :]
                 else:
                     f = cwd + b'/' + f
                 return back + lfutil.standin(f)
 
         else:
 
             def tostandin(f):
                 if lfutil.isstandin(f):
                     return f
                 return lfutil.standin(f)
 
         pats.update(fixpats(f, tostandin) for f in p)
 
         for i in range(0, len(m._files)):
             # Don't add '.hglf' to m.files, since that is already covered by '.'
             if m._files[i] == b'.':
                 continue
             standin = lfutil.standin(m._files[i])
             # If the "standin" is a directory, append instead of replace to
             # support naming a directory on the command line with only
             # largefiles.  The original directory is kept to support normal
             # files.
             if standin in ctx:
                 m._files[i] = standin
             elif m._files[i] not in ctx and repo.wvfs.isdir(standin):
                 m._files.append(standin)
 
         m._fileset = set(m._files)
         m.always = lambda: False
         origmatchfn = m.matchfn
 
         def lfmatchfn(f):
             lf = lfutil.splitstandin(f)
             if lf is not None and origmatchfn(lf):
                 return True
             r = origmatchfn(f)
             return r
 
         m.matchfn = lfmatchfn
 
         ui.debug(b'updated patterns: %s\n' % b', '.join(sorted(pats)))
         return m, pats
 
     # For hg log --patch, the match object is used in two different senses:
     # (1) to determine what revisions should be printed out, and
     # (2) to determine what files to print out diffs for.
     # The magic matchandpats override should be used for case (1) but not for
     # case (2).
     oldmatchandpats = scmutil.matchandpats
 
     def overridemakefilematcher(orig, repo, pats, opts, badfn=None):
         wctx = repo[None]
         match, pats = oldmatchandpats(wctx, pats, opts, badfn=badfn)
         return lambda ctx: match
 
     wrappedmatchandpats = extensions.wrappedfunction(
         scmutil, b'matchandpats', overridematchandpats
     )
     wrappedmakefilematcher = extensions.wrappedfunction(
         logcmdutil, b'_makenofollowfilematcher', overridemakefilematcher
     )
     with wrappedmatchandpats, wrappedmakefilematcher:
         return orig(ui, repo, *pats, **opts)
 
 
 @eh.wrapcommand(
     b'verify',
     opts=[
         (
             b'',
             b'large',
             None,
             _(b'verify that all largefiles in current revision exists'),
         ),
         (
             b'',
             b'lfa',
             None,
             _(b'verify largefiles in all revisions, not just current'),
         ),
         (
             b'',
             b'lfc',
             None,
             _(b'verify local largefile contents, not just existence'),
         ),
     ],
 )
 def overrideverify(orig, ui, repo, *pats, **opts):
     large = opts.pop('large', False)
     all = opts.pop('lfa', False)
     contents = opts.pop('lfc', False)
 
     result = orig(ui, repo, *pats, **opts)
     if large or all or contents:
         result = result or lfcommands.verifylfiles(ui, repo, all, contents)
     return result
 
 
 @eh.wrapcommand(
     b'debugstate',
     opts=[(b'', b'large', None, _(b'display largefiles dirstate'))],
 )
 def overridedebugstate(orig, ui, repo, *pats, **opts):
     large = opts.pop('large', False)
     if large:
 
         class fakerepo:
             dirstate = lfutil.openlfdirstate(ui, repo)
 
         orig(ui, fakerepo, *pats, **opts)
     else:
         orig(ui, repo, *pats, **opts)
 
 
 # Before starting the manifest merge, merge.updates will call
 # _checkunknownfile to check if there are any files in the merged-in
 # changeset that collide with unknown files in the working copy.
 #
 # The largefiles are seen as unknown, so this prevents us from merging
 # in a file 'foo' if we already have a largefile with the same name.
 #
 # The overridden function filters the unknown files by removing any
 # largefiles. This makes the merge proceed and we can then handle this
 # case further in the overridden calculateupdates function below.
 @eh.wrapfunction(merge, b'_checkunknownfile')
 def overridecheckunknownfile(
     origfn, dirstate, wvfs, dircache, wctx, mctx, f, f2=None
 ):
     if lfutil.standin(dirstate.normalize(f)) in wctx:
         return False
     return origfn(dirstate, wvfs, dircache, wctx, mctx, f, f2)
 
 
 # The manifest merge handles conflicts on the manifest level. We want
 # to handle changes in largefile-ness of files at this level too.
 #
 # The strategy is to run the original calculateupdates and then process
 # the action list it outputs. There are two cases we need to deal with:
 #
 # 1. Normal file in p1, largefile in p2. Here the largefile is
 #    detected via its standin file, which will enter the working copy
 #    with a "get" action. It is not "merge" since the standin is all
 #    Mercurial is concerned with at this level -- the link to the
 #    existing normal file is not relevant here.
 #
 # 2. Largefile in p1, normal file in p2. Here we get a "merge" action
 #    since the largefile will be present in the working copy and
 #    different from the normal file in p2. Mercurial therefore
 #    triggers a merge action.
 #
 # In both cases, we prompt the user and emit new actions to either
 # remove the standin (if the normal file was kept) or to remove the
 # normal file and get the standin (if the largefile was kept). The
 # default prompt answer is to use the largefile version since it was
 # presumably changed on purpose.
 #
 # Finally, the merge.applyupdates function will then take care of
 # writing the files into the working copy and lfcommands.updatelfiles
 # will update the largefiles.
 @eh.wrapfunction(merge, b'calculateupdates')
 def overridecalculateupdates(
     origfn, repo, p1, p2, pas, branchmerge, force, acceptremote, *args, **kwargs
 ):
     overwrite = force and not branchmerge
     mresult = origfn(
         repo, p1, p2, pas, branchmerge, force, acceptremote, *args, **kwargs
     )
 
     if overwrite:
         return mresult
 
     # Convert to dictionary with filename as key and action as value.
     lfiles = set()
     for f in mresult.files():
         splitstandin = lfutil.splitstandin(f)
         if splitstandin is not None and splitstandin in p1:
             lfiles.add(splitstandin)
         elif lfutil.standin(f) in p1:
             lfiles.add(f)
 
     for lfile in sorted(lfiles):
         standin = lfutil.standin(lfile)
         (lm, largs, lmsg) = mresult.getfile(lfile, (None, None, None))
         (sm, sargs, smsg) = mresult.getfile(standin, (None, None, None))
 
         if sm in (ACTION_GET, ACTION_DELETED_CHANGED) and lm != ACTION_REMOVE:
             if sm == ACTION_DELETED_CHANGED:
                 f1, f2, fa, move, anc = sargs
                 sargs = (p2[f2].flags(), False)
             # Case 1: normal file in the working copy, largefile in
             # the second parent
             usermsg = (
                 _(
                     b'remote turned local normal file %s into a largefile\n'
                     b'use (l)argefile or keep (n)ormal file?'
                     b'$$ &Largefile $$ &Normal file'
                 )
                 % lfile
             )
             if repo.ui.promptchoice(usermsg, 0) == 0:  # pick remote largefile
                 mresult.addfile(
                     lfile, ACTION_REMOVE, None, b'replaced by standin'
                 )
                 mresult.addfile(standin, ACTION_GET, sargs, b'replaces standin')
             else:  # keep local normal file
                 mresult.addfile(lfile, ACTION_KEEP, None, b'replaces standin')
                 if branchmerge:
                     mresult.addfile(
                         standin,
                         ACTION_KEEP,
                         None,
                         b'replaced by non-standin',
                     )
                 else:
                     mresult.addfile(
                         standin,
                         ACTION_REMOVE,
                         None,
                         b'replaced by non-standin',
                     )
         if lm in (ACTION_GET, ACTION_DELETED_CHANGED) and sm != ACTION_REMOVE:
             if lm == ACTION_DELETED_CHANGED:
                 f1, f2, fa, move, anc = largs
                 largs = (p2[f2].flags(), False)
             # Case 2: largefile in the working copy, normal file in
             # the second parent
             usermsg = (
                 _(
                     b'remote turned local largefile %s into a normal file\n'
                     b'keep (l)argefile or use (n)ormal file?'
                     b'$$ &Largefile $$ &Normal file'
                 )
                 % lfile
             )
             if repo.ui.promptchoice(usermsg, 0) == 0:  # keep local largefile
                 if branchmerge:
                     # largefile can be restored from standin safely
                     mresult.addfile(
                         lfile,
                         ACTION_KEEP,
                         None,
                         b'replaced by standin',
                     )
                     mresult.addfile(
                         standin, ACTION_KEEP, None, b'replaces standin'
                     )
                 else:
                     # "lfile" should be marked as "removed" without
                     # removal of itself
                     mresult.addfile(
                         lfile,
                         MERGE_ACTION_LARGEFILE_MARK_REMOVED,
                         None,
                         b'forget non-standin largefile',
                     )
 
                     # linear-merge should treat this largefile as 're-added'
                     mresult.addfile(standin, ACTION_ADD, None, b'keep standin')
             else:  # pick remote normal file
                 mresult.addfile(lfile, ACTION_GET, largs, b'replaces standin')
                 mresult.addfile(
                     standin,
                     ACTION_REMOVE,
                     None,
                     b'replaced by non-standin',
                 )
 
     return mresult
 
 
 @eh.wrapfunction(mergestatemod, b'recordupdates')
 def mergerecordupdates(orig, repo, actions, branchmerge, getfiledata):
     if MERGE_ACTION_LARGEFILE_MARK_REMOVED in actions:
         lfdirstate = lfutil.openlfdirstate(repo.ui, repo)
         with lfdirstate.changing_parents(repo):
             for lfile, args, msg in actions[
                 MERGE_ACTION_LARGEFILE_MARK_REMOVED
             ]:
                 # this should be executed before 'orig', to execute 'remove'
                 # before all other actions
                 repo.dirstate.update_file(
                     lfile, p1_tracked=True, wc_tracked=False
                 )
                 # make sure lfile doesn't get synclfdirstate'd as normal
                 lfdirstate.update_file(lfile, p1_tracked=False, wc_tracked=True)
         lfdirstate.write(repo.currenttransaction())
 
     return orig(repo, actions, branchmerge, getfiledata)
 
 
 # Override filemerge to prompt the user about how they wish to merge
 # largefiles. This will handle identical edits without prompting the user.
 @eh.wrapfunction(filemerge, b'filemerge')
 def overridefilemerge(
     origfn, repo, wctx, mynode, orig, fcd, fco, fca, labels=None
 ):
     if not lfutil.isstandin(orig) or fcd.isabsent() or fco.isabsent():
         return origfn(repo, wctx, mynode, orig, fcd, fco, fca, labels=labels)
 
     ahash = lfutil.readasstandin(fca).lower()
     dhash = lfutil.readasstandin(fcd).lower()
     ohash = lfutil.readasstandin(fco).lower()
     if (
         ohash != ahash
         and ohash != dhash
         and (
             dhash == ahash
             or repo.ui.promptchoice(
                 _(
                     b'largefile %s has a merge conflict\nancestor was %s\n'
                     b'you can keep (l)ocal %s or take (o)ther %s.\n'
                     b'what do you want to do?'
                     b'$$ &Local $$ &Other'
                 )
                 % (lfutil.splitstandin(orig), ahash, dhash, ohash),
                 0,
             )
             == 1
         )
     ):
         repo.wwrite(fcd.path(), fco.data(), fco.flags())
     return 0, False
 
 
 @eh.wrapfunction(copiesmod, b'pathcopies')
 def copiespathcopies(orig, ctx1, ctx2, match=None):
     copies = orig(ctx1, ctx2, match=match)
     updated = {}
 
     for k, v in copies.items():
         updated[lfutil.splitstandin(k) or k] = lfutil.splitstandin(v) or v
 
     return updated
 
 
 # Copy first changes the matchers to match standins instead of
 # largefiles.  Then it overrides util.copyfile in that function it
 # checks if the destination largefile already exists. It also keeps a
 # list of copied files so that the largefiles can be copied and the
 # dirstate updated.
 @eh.wrapfunction(cmdutil, b'copy')
 def overridecopy(orig, ui, repo, pats, opts, rename=False):
     # doesn't remove largefile on rename
     if len(pats) < 2:
         # this isn't legal, let the original function deal with it
         return orig(ui, repo, pats, opts, rename)
 
     # This could copy both lfiles and normal files in one command,
     # but we don't want to do that. First replace their matcher to
     # only match normal files and run it, then replace it to just
     # match largefiles and run it again.
     nonormalfiles = False
     nolfiles = False
     manifest = repo[None].manifest()
 
     def normalfilesmatchfn(
         orig,
         ctx,
         pats=(),
         opts=None,
         globbed=False,
         default=b'relpath',
         badfn=None,
     ):
         if opts is None:
             opts = {}
         match = orig(ctx, pats, opts, globbed, default, badfn=badfn)
         return composenormalfilematcher(match, manifest)
 
     with extensions.wrappedfunction(scmutil, b'match', normalfilesmatchfn):
         try:
             result = orig(ui, repo, pats, opts, rename)
         except error.Abort as e:
             if e.message != _(b'no files to copy'):
                 raise e
             else:
                 nonormalfiles = True
             result = 0
 
     # The first rename can cause our current working directory to be removed.
     # In that case there is nothing left to copy/rename so just quit.
     try:
         repo.getcwd()
     except OSError:
         return result
 
     def makestandin(relpath):
         path = pathutil.canonpath(repo.root, repo.getcwd(), relpath)
         return repo.wvfs.join(lfutil.standin(path))
 
     fullpats = scmutil.expandpats(pats)
     dest = fullpats[-1]
 
     if os.path.isdir(dest):
         if not os.path.isdir(makestandin(dest)):
             os.makedirs(makestandin(dest))
 
     try:
         # When we call orig below it creates the standins but we don't add
         # them to the dir state until later so lock during that time.
         wlock = repo.wlock()
 
         manifest = repo[None].manifest()
 
         def overridematch(
             orig,
             ctx,
             pats=(),
             opts=None,
             globbed=False,
             default=b'relpath',
             badfn=None,
         ):
             if opts is None:
                 opts = {}
             newpats = []
             # The patterns were previously mangled to add the standin
             # directory; we need to remove that now
             for pat in pats:
                 if matchmod.patkind(pat) is None and lfutil.shortname in pat:
                     newpats.append(pat.replace(lfutil.shortname, b''))
                 else:
                     newpats.append(pat)
             match = orig(ctx, newpats, opts, globbed, default, badfn=badfn)
             m = copy.copy(match)
             lfile = lambda f: lfutil.standin(f) in manifest
             m._files = [lfutil.standin(f) for f in m._files if lfile(f)]
             m._fileset = set(m._files)
             origmatchfn = m.matchfn
 
             def matchfn(f):
                 lfile = lfutil.splitstandin(f)
                 return (
                     lfile is not None
                     and (f in manifest)
                     and origmatchfn(lfile)
                     or None
                 )
 
             m.matchfn = matchfn
             return m
 
         listpats = []
         for pat in pats:
             if matchmod.patkind(pat) is not None:
                 listpats.append(pat)
             else:
                 listpats.append(makestandin(pat))
 
         copiedfiles = []
 
         def overridecopyfile(orig, src, dest, *args, **kwargs):
             if lfutil.shortname in src and dest.startswith(
                 repo.wjoin(lfutil.shortname)
             ):
                 destlfile = dest.replace(lfutil.shortname, b'')
                 if not opts[b'force'] and os.path.exists(destlfile):
                     raise IOError(
                         b'', _(b'destination largefile already exists')
                     )
             copiedfiles.append((src, dest))
             orig(src, dest, *args, **kwargs)
 
         with extensions.wrappedfunction(util, b'copyfile', overridecopyfile):
             with extensions.wrappedfunction(scmutil, b'match', overridematch):
                 result += orig(ui, repo, listpats, opts, rename)
 
         lfdirstate = lfutil.openlfdirstate(ui, repo)
         for (src, dest) in copiedfiles:
             if lfutil.shortname in src and dest.startswith(
                 repo.wjoin(lfutil.shortname)
             ):
                 srclfile = src.replace(repo.wjoin(lfutil.standin(b'')), b'')
                 destlfile = dest.replace(repo.wjoin(lfutil.standin(b'')), b'')
                 destlfiledir = repo.wvfs.dirname(repo.wjoin(destlfile)) or b'.'
                 if not os.path.isdir(destlfiledir):
                     os.makedirs(destlfiledir)
                 if rename:
                     os.rename(repo.wjoin(srclfile), repo.wjoin(destlfile))
 
                     # The file is gone, but this deletes any empty parent
                     # directories as a side-effect.
                     repo.wvfs.unlinkpath(srclfile, ignoremissing=True)
                     lfdirstate.set_untracked(srclfile)
                 else:
                     util.copyfile(repo.wjoin(srclfile), repo.wjoin(destlfile))
 
                 lfdirstate.set_tracked(destlfile)
         lfdirstate.write(repo.currenttransaction())
     except error.Abort as e:
         if e.message != _(b'no files to copy'):
             raise e
         else:
             nolfiles = True
     finally:
         wlock.release()
 
     if nolfiles and nonormalfiles:
         raise error.Abort(_(b'no files to copy'))
 
     return result
 
 
 # When the user calls revert, we have to be careful to not revert any
 # changes to other largefiles accidentally. This means we have to keep
 # track of the largefiles that are being reverted so we only pull down
 # the necessary largefiles.
 #
 # Standins are only updated (to match the hash of largefiles) before
 # commits. Update the standins then run the original revert, changing
 # the matcher to hit standins instead of largefiles. Based on the
 # resulting standins update the largefiles.
 @eh.wrapfunction(cmdutil, b'revert')
 def overriderevert(orig, ui, repo, ctx, *pats, **opts):
     # Because we put the standins in a bad state (by updating them)
     # and then return them to a correct state we need to lock to
     # prevent others from changing them in their incorrect state.
     with repo.wlock():
         lfdirstate = lfutil.openlfdirstate(ui, repo)
         s = lfutil.lfdirstatestatus(lfdirstate, repo)
         lfdirstate.write(repo.currenttransaction())
         for lfile in s.modified:
             lfutil.updatestandin(repo, lfile, lfutil.standin(lfile))
         for lfile in s.deleted:
             fstandin = lfutil.standin(lfile)
             if repo.wvfs.exists(fstandin):
                 repo.wvfs.unlink(fstandin)
 
         oldstandins = lfutil.getstandinsstate(repo)
 
         def overridematch(
             orig,
             mctx,
             pats=(),
             opts=None,
             globbed=False,
             default=b'relpath',
             badfn=None,
         ):
             if opts is None:
                 opts = {}
             match = orig(mctx, pats, opts, globbed, default, badfn=badfn)
             m = copy.copy(match)
 
             # revert supports recursing into subrepos, and though largefiles
             # currently doesn't work correctly in that case, this match is
             # called, so the lfdirstate above may not be the correct one for
             # this invocation of match.
             lfdirstate = lfutil.openlfdirstate(
                 mctx.repo().ui, mctx.repo(), False
             )
 
             wctx = repo[None]
             matchfiles = []
             for f in m._files:
                 standin = lfutil.standin(f)
                 if standin in ctx or standin in mctx:
                     matchfiles.append(standin)
                 elif standin in wctx or lfdirstate.get_entry(f).removed:
                     continue
                 else:
                     matchfiles.append(f)
             m._files = matchfiles
             m._fileset = set(m._files)
             origmatchfn = m.matchfn
 
             def matchfn(f):
                 lfile = lfutil.splitstandin(f)
                 if lfile is not None:
                     return origmatchfn(lfile) and (f in ctx or f in mctx)
                 return origmatchfn(f)
 
             m.matchfn = matchfn
             return m
 
         with extensions.wrappedfunction(scmutil, b'match', overridematch):
             orig(ui, repo, ctx, *pats, **opts)
 
         newstandins = lfutil.getstandinsstate(repo)
         filelist = lfutil.getlfilestoupdate(oldstandins, newstandins)
         # lfdirstate should be 'normallookup'-ed for updated files,
         # because reverting doesn't touch dirstate for 'normal' files
         # when target revision is explicitly specified: in such case,
         # 'n' and valid timestamp in dirstate doesn't ensure 'clean'
         # of target (standin) file.
         lfcommands.updatelfiles(
             ui, repo, filelist, printmessage=False, normallookup=True
         )
 
 
 # after pulling changesets, we need to take some extra care to get
 # largefiles updated remotely
 @eh.wrapcommand(
     b'pull',
     opts=[
         (
             b'',
             b'all-largefiles',
             None,
             _(b'download all pulled versions of largefiles (DEPRECATED)'),
         ),
         (
             b'',
             b'lfrev',
             [],
             _(b'download largefiles for these revisions'),
             _(b'REV'),
         ),
     ],
 )
 def overridepull(orig, ui, repo, source=None, **opts):
     revsprepull = len(repo)
     if not source:
         source = b'default'
     repo.lfpullsource = source
     result = orig(ui, repo, source, **opts)
     revspostpull = len(repo)
     lfrevs = opts.get('lfrev', [])
     if opts.get('all_largefiles'):
         lfrevs.append(b'pulled()')
     if lfrevs and revspostpull > revsprepull:
         numcached = 0
         repo.firstpulled = revsprepull  # for pulled() revset expression
         try:
             for rev in logcmdutil.revrange(repo, lfrevs):
                 ui.note(_(b'pulling largefiles for revision %d\n') % rev)
                 (cached, missing) = lfcommands.cachelfiles(ui, repo, rev)
                 numcached += len(cached)
         finally:
             del repo.firstpulled
         ui.status(_(b"%d largefiles cached\n") % numcached)
     return result
 
 
 @eh.wrapcommand(
     b'push',
     opts=[
         (
             b'',
             b'lfrev',
             [],
             _(b'upload largefiles for these revisions'),
             _(b'REV'),
         )
     ],
 )
 def overridepush(orig, ui, repo, *args, **kwargs):
     """Override push command and store --lfrev parameters in opargs"""
     lfrevs = kwargs.pop('lfrev', None)
     if lfrevs:
         opargs = kwargs.setdefault('opargs', {})
         opargs[b'lfrevs'] = logcmdutil.revrange(repo, lfrevs)
     return orig(ui, repo, *args, **kwargs)
 
 
 @eh.wrapfunction(exchange, b'pushoperation')
 def exchangepushoperation(orig, *args, **kwargs):
     """Override pushoperation constructor and store lfrevs parameter"""
     lfrevs = kwargs.pop('lfrevs', None)
     pushop = orig(*args, **kwargs)
     pushop.lfrevs = lfrevs
     return pushop
 
 
 @eh.revsetpredicate(b'pulled()')
 def pulledrevsetsymbol(repo, subset, x):
     """Changesets that just has been pulled.
 
     Only available with largefiles from pull --lfrev expressions.
 
     .. container:: verbose
 
       Some examples:
 
       - pull largefiles for all new changesets::
 
           hg pull -lfrev "pulled()"
 
       - pull largefiles for all new branch heads::
 
           hg pull -lfrev "head(pulled()) and not closed()"
 
     """
 
     try:
         firstpulled = repo.firstpulled
     except AttributeError:
         raise error.Abort(_(b"pulled() only available in --lfrev"))
     return smartset.baseset([r for r in subset if r >= firstpulled])
 
 
 @eh.wrapcommand(
     b'clone',
     opts=[
         (
             b'',
             b'all-largefiles',
             None,
             _(b'download all versions of all largefiles'),
         )
     ],
 )
 def overrideclone(orig, ui, source, dest=None, **opts):
     d = dest
     if d is None:
         d = hg.defaultdest(source)
     if opts.get('all_largefiles') and not hg.islocal(d):
         raise error.Abort(
             _(b'--all-largefiles is incompatible with non-local destination %s')
             % d
         )
 
     return orig(ui, source, dest, **opts)
 
 
 @eh.wrapfunction(hg, b'clone')
 def hgclone(orig, ui, opts, *args, **kwargs):
     result = orig(ui, opts, *args, **kwargs)
 
     if result is not None:
         sourcerepo, destrepo = result
         repo = destrepo.local()
 
         # When cloning to a remote repo (like through SSH), no repo is available
         # from the peer.   Therefore the largefiles can't be downloaded and the
         # hgrc can't be updated.
         if not repo:
             return result
 
         # Caching is implicitly limited to 'rev' option, since the dest repo was
         # truncated at that point.  The user may expect a download count with
         # this option, so attempt whether or not this is a largefile repo.
         if opts.get(b'all_largefiles'):
             success, missing = lfcommands.downloadlfiles(ui, repo)
 
             if missing != 0:
                 return None
 
     return result
 
 
 @eh.wrapcommand(b'rebase', extension=b'rebase')
 def overriderebasecmd(orig, ui, repo, **opts):
     if not util.safehasattr(repo, b'_largefilesenabled'):
         return orig(ui, repo, **opts)
 
     resuming = opts.get('continue')
     repo._lfcommithooks.append(lfutil.automatedcommithook(resuming))
     repo._lfstatuswriters.append(lambda *msg, **opts: None)
     try:
         with ui.configoverride(
             {(b'rebase', b'experimental.inmemory'): False}, b"largefiles"
         ):
             return orig(ui, repo, **opts)
     finally:
         repo._lfstatuswriters.pop()
         repo._lfcommithooks.pop()
 
 
 @eh.extsetup
 def overriderebase(ui):
     try:
         rebase = extensions.find(b'rebase')
     except KeyError:
         pass
     else:
 
         def _dorebase(orig, *args, **kwargs):
             kwargs['inmemory'] = False
             return orig(*args, **kwargs)
 
         extensions.wrapfunction(rebase, b'_dorebase', _dorebase)
 
 
 @eh.wrapcommand(b'archive')
 def overridearchivecmd(orig, ui, repo, dest, **opts):
     with lfstatus(repo.unfiltered()):
         return orig(ui, repo.unfiltered(), dest, **opts)
 
 
 @eh.wrapfunction(webcommands, b'archive')
 def hgwebarchive(orig, web):
     with lfstatus(web.repo):
         return orig(web)
 
 
 @eh.wrapfunction(archival, b'archive')
 def overridearchive(
     orig,
     repo,
     dest,
     node,
     kind,
     decode=True,
     match=None,
     prefix=b'',
     mtime=None,
     subrepos=None,
 ):
     # For some reason setting repo.lfstatus in hgwebarchive only changes the
     # unfiltered repo's attr, so check that as well.
     if not repo.lfstatus and not repo.unfiltered().lfstatus:
         return orig(
             repo, dest, node, kind, decode, match, prefix, mtime, subrepos
         )
 
     # No need to lock because we are only reading history and
     # largefile caches, neither of which are modified.
     if node is not None:
         lfcommands.cachelfiles(repo.ui, repo, node)
 
     if kind not in archival.archivers:
         raise error.Abort(_(b"unknown archive type '%s'") % kind)
 
     ctx = repo[node]
 
     if kind == b'files':
         if prefix:
             raise error.Abort(_(b'cannot give prefix when archiving to files'))
     else:
         prefix = archival.tidyprefix(dest, kind, prefix)
 
     def write(name, mode, islink, getdata):
         if match and not match(name):
             return
         data = getdata()
         if decode:
             data = repo.wwritedata(name, data)
         archiver.addfile(prefix + name, mode, islink, data)
 
     archiver = archival.archivers[kind](dest, mtime or ctx.date()[0])
 
     if repo.ui.configbool(b"ui", b"archivemeta"):
         write(
             b'.hg_archival.txt',
             0o644,
             False,
             lambda: archival.buildmetadata(ctx),
         )
 
     for f in ctx:
         ff = ctx.flags(f)
         getdata = ctx[f].data
         lfile = lfutil.splitstandin(f)
         if lfile is not None:
             if node is not None:
                 path = lfutil.findfile(repo, getdata().strip())
 
                 if path is None:
                     raise error.Abort(
                         _(
                             b'largefile %s not found in repo store or system cache'
                         )
                         % lfile
                     )
             else:
                 path = lfile
 
             f = lfile
 
             getdata = lambda: util.readfile(path)
         write(f, b'x' in ff and 0o755 or 0o644, b'l' in ff, getdata)
 
     if subrepos:
         for subpath in sorted(ctx.substate):
             sub = ctx.workingsub(subpath)
             submatch = matchmod.subdirmatcher(subpath, match)
             subprefix = prefix + subpath + b'/'
 
             # TODO: Only hgsubrepo instances have `_repo`, so figure out how to
             # infer and possibly set lfstatus in hgsubrepoarchive.  That would
             # allow only hgsubrepos to set this, instead of the current scheme
             # where the parent sets this for the child.
             with (
                 util.safehasattr(sub, '_repo')
                 and lfstatus(sub._repo)
                 or util.nullcontextmanager()
             ):
                 sub.archive(archiver, subprefix, submatch)
 
     archiver.done()
 
 
 @eh.wrapfunction(subrepo.hgsubrepo, b'archive')
 def hgsubrepoarchive(orig, repo, archiver, prefix, match=None, decode=True):
     lfenabled = util.safehasattr(repo._repo, b'_largefilesenabled')
     if not lfenabled or not repo._repo.lfstatus:
         return orig(repo, archiver, prefix, match, decode)
 
     repo._get(repo._state + (b'hg',))
     rev = repo._state[1]
     ctx = repo._repo[rev]
 
     if ctx.node() is not None:
         lfcommands.cachelfiles(repo.ui, repo._repo, ctx.node())
 
     def write(name, mode, islink, getdata):
         # At this point, the standin has been replaced with the largefile name,
         # so the normal matcher works here without the lfutil variants.
         if match and not match(f):
             return
         data = getdata()
         if decode:
             data = repo._repo.wwritedata(name, data)
 
         archiver.addfile(prefix + name, mode, islink, data)
 
     for f in ctx:
         ff = ctx.flags(f)
         getdata = ctx[f].data
         lfile = lfutil.splitstandin(f)
         if lfile is not None:
             if ctx.node() is not None:
                 path = lfutil.findfile(repo._repo, getdata().strip())
 
                 if path is None:
                     raise error.Abort(
                         _(
                             b'largefile %s not found in repo store or system cache'
                         )
                         % lfile
                     )
             else:
                 path = lfile
 
             f = lfile
 
             getdata = lambda: util.readfile(os.path.join(prefix, path))
 
         write(f, b'x' in ff and 0o755 or 0o644, b'l' in ff, getdata)
 
     for subpath in sorted(ctx.substate):
         sub = ctx.workingsub(subpath)
         submatch = matchmod.subdirmatcher(subpath, match)
         subprefix = prefix + subpath + b'/'
         # TODO: Only hgsubrepo instances have `_repo`, so figure out how to
         # infer and possibly set lfstatus at the top of this function.  That
         # would allow only hgsubrepos to set this, instead of the current scheme
         # where the parent sets this for the child.
         with (
             util.safehasattr(sub, '_repo')
             and lfstatus(sub._repo)
             or util.nullcontextmanager()
         ):
             sub.archive(archiver, subprefix, submatch, decode)
 
 
 # If a largefile is modified, the change is not reflected in its
 # standin until a commit. cmdutil.bailifchanged() raises an exception
 # if the repo has uncommitted changes. Wrap it to also check if
 # largefiles were changed. This is used by bisect, backout and fetch.
 @eh.wrapfunction(cmdutil, b'bailifchanged')
 def overridebailifchanged(orig, repo, *args, **kwargs):
     orig(repo, *args, **kwargs)
     with lfstatus(repo):
         s = repo.status()
     if s.modified or s.added or s.removed or s.deleted:
         raise error.Abort(_(b'uncommitted changes'))
 
 
 @eh.wrapfunction(cmdutil, b'postcommitstatus')
 def postcommitstatus(orig, repo, *args, **kwargs):
     with lfstatus(repo):
         return orig(repo, *args, **kwargs)
 
 
 @eh.wrapfunction(cmdutil, b'forget')
 def cmdutilforget(
     orig, ui, repo, match, prefix, uipathfn, explicitonly, dryrun, interactive
 ):
     normalmatcher = composenormalfilematcher(match, repo[None].manifest())
     bad, forgot = orig(
         ui,
         repo,
         normalmatcher,
         prefix,
         uipathfn,
         explicitonly,
         dryrun,
         interactive,
     )
     m = composelargefilematcher(match, repo[None].manifest())
 
     with lfstatus(repo):
         s = repo.status(match=m, clean=True)
     manifest = repo[None].manifest()
     forget = sorted(s.modified + s.added + s.deleted + s.clean)
     forget = [f for f in forget if lfutil.standin(f) in manifest]
 
     for f in forget:
         fstandin = lfutil.standin(f)
         if fstandin not in repo.dirstate and not repo.wvfs.isdir(fstandin):
             ui.warn(
                 _(b'not removing %s: file is already untracked\n') % uipathfn(f)
             )
             bad.append(f)
 
     for f in forget:
         if ui.verbose or not m.exact(f):
             ui.status(_(b'removing %s\n') % uipathfn(f))
 
     # Need to lock because standin files are deleted then removed from the
     # repository and we could race in-between.
     with repo.wlock():
         lfdirstate = lfutil.openlfdirstate(ui, repo)
         for f in forget:
             lfdirstate.set_untracked(f)
         lfdirstate.write(repo.currenttransaction())
         standins = [lfutil.standin(f) for f in forget]
         for f in standins:
             repo.wvfs.unlinkpath(f, ignoremissing=True)
         rejected = repo[None].forget(standins)
 
     bad.extend(f for f in rejected if f in m.files())
     forgot.extend(f for f in forget if f not in rejected)
     return bad, forgot
 
 
 def _getoutgoings(repo, other, missing, addfunc):
     """get pairs of filename and largefile hash in outgoing revisions
     in 'missing'.
 
     largefiles already existing on 'other' repository are ignored.
 
     'addfunc' is invoked with each unique pairs of filename and
     largefile hash value.
     """
     knowns = set()
     lfhashes = set()
 
     def dedup(fn, lfhash):
         k = (fn, lfhash)
         if k not in knowns:
             knowns.add(k)
             lfhashes.add(lfhash)
 
     lfutil.getlfilestoupload(repo, missing, dedup)
     if lfhashes:
         lfexists = storefactory.openstore(repo, other).exists(lfhashes)
         for fn, lfhash in knowns:
             if not lfexists[lfhash]:  # lfhash doesn't exist on "other"
                 addfunc(fn, lfhash)
 
 
 def outgoinghook(ui, repo, other, opts, missing):
     if opts.pop(b'large', None):
         lfhashes = set()
         if ui.debugflag:
             toupload = {}
 
             def addfunc(fn, lfhash):
                 if fn not in toupload:
                     toupload[fn] = []  # pytype: disable=unsupported-operands
                 toupload[fn].append(lfhash)
                 lfhashes.add(lfhash)
 
             def showhashes(fn):
                 for lfhash in sorted(toupload[fn]):
                     ui.debug(b'    %s\n' % lfhash)
 
         else:
             toupload = set()
 
             def addfunc(fn, lfhash):
                 toupload.add(fn)
                 lfhashes.add(lfhash)
 
             def showhashes(fn):
                 pass
 
         _getoutgoings(repo, other, missing, addfunc)
 
         if not toupload:
             ui.status(_(b'largefiles: no files to upload\n'))
         else:
             ui.status(
                 _(b'largefiles to upload (%d entities):\n') % (len(lfhashes))
             )
             for file in sorted(toupload):
                 ui.status(lfutil.splitstandin(file) + b'\n')
                 showhashes(file)
             ui.status(b'\n')
 
 
 @eh.wrapcommand(
     b'outgoing', opts=[(b'', b'large', None, _(b'display outgoing largefiles'))]
 )
 def _outgoingcmd(orig, *args, **kwargs):
     # Nothing to do here other than add the extra help option- the hook above
     # processes it.
     return orig(*args, **kwargs)
 
 
 def summaryremotehook(ui, repo, opts, changes):
     largeopt = opts.get(b'large', False)
     if changes is None:
         if largeopt:
             return (False, True)  # only outgoing check is needed
         else:
             return (False, False)
     elif largeopt:
         url, branch, peer, outgoing = changes[1]
         if peer is None:
             # i18n: column positioning for "hg summary"
             ui.status(_(b'largefiles: (no remote repo)\n'))
             return
 
         toupload = set()
         lfhashes = set()
 
         def addfunc(fn, lfhash):
             toupload.add(fn)
             lfhashes.add(lfhash)
 
         _getoutgoings(repo, peer, outgoing.missing, addfunc)
 
         if not toupload:
             # i18n: column positioning for "hg summary"
             ui.status(_(b'largefiles: (no files to upload)\n'))
         else:
             # i18n: column positioning for "hg summary"
             ui.status(
                 _(b'largefiles: %d entities for %d files to upload\n')
                 % (len(lfhashes), len(toupload))
             )
 
 
 @eh.wrapcommand(
     b'summary', opts=[(b'', b'large', None, _(b'display outgoing largefiles'))]
 )
 def overridesummary(orig, ui, repo, *pats, **opts):
     with lfstatus(repo):
         orig(ui, repo, *pats, **opts)
 
 
 @eh.wrapfunction(scmutil, b'addremove')
-def scmutiladdremove(orig, repo, matcher, prefix, uipathfn, opts=None):
+def scmutiladdremove(
+    orig,
+    repo,
+    matcher,
+    prefix,
+    uipathfn,
+    opts=None,
+    open_tr=None,
+):
     if opts is None:
         opts = {}
     if not lfutil.islfilesrepo(repo):
-        return orig(repo, matcher, prefix, uipathfn, opts)
+        return orig(repo, matcher, prefix, uipathfn, opts, open_tr=open_tr)
     # Get the list of missing largefiles so we can remove them
     lfdirstate = lfutil.openlfdirstate(repo.ui, repo)
     unsure, s, mtime_boundary = lfdirstate.status(
         matchmod.always(),
         subrepos=[],
         ignored=False,
         clean=False,
         unknown=False,
     )
 
+    # open the transaction and changing_files context
+    if open_tr is not None:
+        open_tr()
+
     # Call into the normal remove code, but the removing of the standin, we want
     # to have handled by original addremove.  Monkey patching here makes sure
     # we don't remove the standin in the largefiles code, preventing a very
     # confused state later.
     if s.deleted:
         m = copy.copy(matcher)
 
         # The m._files and m._map attributes are not changed to the deleted list
         # because that affects the m.exact() test, which in turn governs whether
         # or not the file name is printed, and how.  Simply limit the original
         # matches to those in the deleted status list.
         matchfn = m.matchfn
         m.matchfn = lambda f: f in s.deleted and matchfn(f)
 
         removelargefiles(
             repo.ui,
             repo,
             True,
             m,
             uipathfn,
             opts.get(b'dry_run'),
             **pycompat.strkwargs(opts)
         )
     # Call into the normal add code, and any files that *should* be added as
     # largefiles will be
     added, bad = addlargefiles(
         repo.ui, repo, True, matcher, uipathfn, **pycompat.strkwargs(opts)
     )
     # Now that we've handled largefiles, hand off to the original addremove
     # function to take care of the rest.  Make sure it doesn't do anything with
     # largefiles by passing a matcher that will ignore them.
     matcher = composenormalfilematcher(matcher, repo[None].manifest(), added)
-    return orig(repo, matcher, prefix, uipathfn, opts)
+
+    return orig(repo, matcher, prefix, uipathfn, opts, open_tr=open_tr)
 
 
 # Calling purge with --all will cause the largefiles to be deleted.
 # Override repo.status to prevent this from happening.
 @eh.wrapcommand(b'purge')
 def overridepurge(orig, ui, repo, *dirs, **opts):
     # XXX Monkey patching a repoview will not work. The assigned attribute will
     # be set on the unfiltered repo, but we will only lookup attributes in the
     # unfiltered repo if the lookup in the repoview object itself fails. As the
     # monkey patched method exists on the repoview class the lookup will not
     # fail. As a result, the original version will shadow the monkey patched
     # one, defeating the monkey patch.
     #
     # As a work around we use an unfiltered repo here. We should do something
     # cleaner instead.
     repo = repo.unfiltered()
     oldstatus = repo.status
 
     def overridestatus(
         node1=b'.',
         node2=None,
         match=None,
         ignored=False,
         clean=False,
         unknown=False,
         listsubrepos=False,
     ):
         r = oldstatus(
             node1, node2, match, ignored, clean, unknown, listsubrepos
         )
         lfdirstate = lfutil.openlfdirstate(ui, repo)
         unknown = [
             f for f in r.unknown if not lfdirstate.get_entry(f).any_tracked
         ]
         ignored = [
             f for f in r.ignored if not lfdirstate.get_entry(f).any_tracked
         ]
         return scmutil.status(
             r.modified, r.added, r.removed, r.deleted, unknown, ignored, r.clean
         )
 
     repo.status = overridestatus
     orig(ui, repo, *dirs, **opts)
     repo.status = oldstatus
 
 
 @eh.wrapcommand(b'rollback')
 def overriderollback(orig, ui, repo, **opts):
     with repo.wlock():
         before = repo.dirstate.parents()
         orphans = {
             f
             for f in repo.dirstate
             if lfutil.isstandin(f) and not repo.dirstate.get_entry(f).removed
         }
         result = orig(ui, repo, **opts)
         after = repo.dirstate.parents()
         if before == after:
             return result  # no need to restore standins
 
         pctx = repo[b'.']
         for f in repo.dirstate:
             if lfutil.isstandin(f):
                 orphans.discard(f)
                 if repo.dirstate.get_entry(f).removed:
                     repo.wvfs.unlinkpath(f, ignoremissing=True)
                 elif f in pctx:
                     fctx = pctx[f]
                     repo.wwrite(f, fctx.data(), fctx.flags())
                 else:
                     # content of standin is not so important in 'a',
                     # 'm' or 'n' (coming from the 2nd parent) cases
                     lfutil.writestandin(repo, f, b'', False)
         for standin in orphans:
             repo.wvfs.unlinkpath(standin, ignoremissing=True)
 
     return result
 
 
 @eh.wrapcommand(b'transplant', extension=b'transplant')
 def overridetransplant(orig, ui, repo, *revs, **opts):
     resuming = opts.get('continue')
     repo._lfcommithooks.append(lfutil.automatedcommithook(resuming))
     repo._lfstatuswriters.append(lambda *msg, **opts: None)
     try:
         result = orig(ui, repo, *revs, **opts)
     finally:
         repo._lfstatuswriters.pop()
         repo._lfcommithooks.pop()
     return result
 
 
 @eh.wrapcommand(b'cat')
 def overridecat(orig, ui, repo, file1, *pats, **opts):
     opts = pycompat.byteskwargs(opts)
     ctx = logcmdutil.revsingle(repo, opts.get(b'rev'))
     err = 1
     notbad = set()
     m = scmutil.match(ctx, (file1,) + pats, opts)
     origmatchfn = m.matchfn
 
     def lfmatchfn(f):
         if origmatchfn(f):
             return True
         lf = lfutil.splitstandin(f)
         if lf is None:
             return False
         notbad.add(lf)
         return origmatchfn(lf)
 
     m.matchfn = lfmatchfn
     origbadfn = m.bad
 
     def lfbadfn(f, msg):
         if not f in notbad:
             origbadfn(f, msg)
 
     m.bad = lfbadfn
 
     origvisitdirfn = m.visitdir
 
     def lfvisitdirfn(dir):
         if dir == lfutil.shortname:
             return True
         ret = origvisitdirfn(dir)
         if ret:
             return ret
         lf = lfutil.splitstandin(dir)
         if lf is None:
             return False
         return origvisitdirfn(lf)
 
     m.visitdir = lfvisitdirfn
 
     for f in ctx.walk(m):
         with cmdutil.makefileobj(ctx, opts.get(b'output'), pathname=f) as fp:
             lf = lfutil.splitstandin(f)
             if lf is None or origmatchfn(f):
                 # duplicating unreachable code from commands.cat
                 data = ctx[f].data()
                 if opts.get(b'decode'):
                     data = repo.wwritedata(f, data)
                 fp.write(data)
             else:
                 hash = lfutil.readasstandin(ctx[f])
                 if not lfutil.inusercache(repo.ui, hash):
                     store = storefactory.openstore(repo)
                     success, missing = store.get([(lf, hash)])
                     if len(success) != 1:
                         raise error.Abort(
                             _(
                                 b'largefile %s is not in cache and could not be '
                                 b'downloaded'
                             )
                             % lf
                         )
                 path = lfutil.usercachepath(repo.ui, hash)
                 with open(path, b"rb") as fpin:
                     for chunk in util.filechunkiter(fpin):
                         fp.write(chunk)
         err = 0
     return err
 
 
 @eh.wrapfunction(merge, b'_update')
 def mergeupdate(orig, repo, node, branchmerge, force, *args, **kwargs):
     matcher = kwargs.get('matcher', None)
     # note if this is a partial update
     partial = matcher and not matcher.always()
     with repo.wlock():
         # branch |       |         |
         #  merge | force | partial | action
         # -------+-------+---------+--------------
         #    x   |   x   |    x    | linear-merge
         #    o   |   x   |    x    | branch-merge
         #    x   |   o   |    x    | overwrite (as clean update)
         #    o   |   o   |    x    | force-branch-merge (*1)
         #    x   |   x   |    o    |   (*)
         #    o   |   x   |    o    |   (*)
         #    x   |   o   |    o    | overwrite (as revert)
         #    o   |   o   |    o    |   (*)
         #
         # (*) don't care
         # (*1) deprecated, but used internally (e.g: "rebase --collapse")
 
         lfdirstate = lfutil.openlfdirstate(repo.ui, repo)
         unsure, s, mtime_boundary = lfdirstate.status(
             matchmod.always(),
             subrepos=[],
             ignored=False,
             clean=True,
             unknown=False,
         )
         oldclean = set(s.clean)
         pctx = repo[b'.']
         dctx = repo[node]
         for lfile in unsure + s.modified:
             lfileabs = repo.wvfs.join(lfile)
             if not repo.wvfs.exists(lfileabs):
                 continue
             lfhash = lfutil.hashfile(lfileabs)
             standin = lfutil.standin(lfile)
             lfutil.writestandin(
                 repo, standin, lfhash, lfutil.getexecutable(lfileabs)
             )
             if standin in pctx and lfhash == lfutil.readasstandin(
                 pctx[standin]
             ):
                 oldclean.add(lfile)
         for lfile in s.added:
             fstandin = lfutil.standin(lfile)
             if fstandin not in dctx:
                 # in this case, content of standin file is meaningless
                 # (in dctx, lfile is unknown, or normal file)
                 continue
             lfutil.updatestandin(repo, lfile, fstandin)
         # mark all clean largefiles as dirty, just in case the update gets
         # interrupted before largefiles and lfdirstate are synchronized
         for lfile in oldclean:
             lfdirstate.set_possibly_dirty(lfile)
         lfdirstate.write(repo.currenttransaction())
 
         oldstandins = lfutil.getstandinsstate(repo)
         wc = kwargs.get('wc')
         if wc and wc.isinmemory():
             # largefiles is not a good candidate for in-memory merge (large
             # files, custom dirstate, matcher usage).
             raise error.ProgrammingError(
                 b'largefiles is not compatible with in-memory merge'
             )
         with lfdirstate.changing_parents(repo):
             result = orig(repo, node, branchmerge, force, *args, **kwargs)
 
             newstandins = lfutil.getstandinsstate(repo)
             filelist = lfutil.getlfilestoupdate(oldstandins, newstandins)
 
             # to avoid leaving all largefiles as dirty and thus rehash them, mark
             # all the ones that didn't change as clean
             for lfile in oldclean.difference(filelist):
                 lfdirstate.update_file(lfile, p1_tracked=True, wc_tracked=True)
             lfdirstate.write(repo.currenttransaction())
 
             if branchmerge or force or partial:
                 filelist.extend(s.deleted + s.removed)
 
             lfcommands.updatelfiles(
                 repo.ui, repo, filelist=filelist, normallookup=partial
             )
 
         return result
 
 
 @eh.wrapfunction(scmutil, b'marktouched')
 def scmutilmarktouched(orig, repo, files, *args, **kwargs):
     result = orig(repo, files, *args, **kwargs)
 
     filelist = []
     for f in files:
         lf = lfutil.splitstandin(f)
         if lf is not None:
             filelist.append(lf)
     if filelist:
         lfcommands.updatelfiles(
             repo.ui,
             repo,
             filelist=filelist,
             printmessage=False,
             normallookup=True,
         )
 
     return result
 
 
 @eh.wrapfunction(upgrade_actions, b'preservedrequirements')
 @eh.wrapfunction(upgrade_actions, b'supporteddestrequirements')
 def upgraderequirements(orig, repo):
     reqs = orig(repo)
     if b'largefiles' in repo.requirements:
         reqs.add(b'largefiles')
     return reqs
 
 
 _lfscheme = b'largefile://'
 
 
 @eh.wrapfunction(urlmod, b'open')
 def openlargefile(orig, ui, url_, data=None, **kwargs):
     if url_.startswith(_lfscheme):
         if data:
             msg = b"cannot use data on a 'largefile://' url"
             raise error.ProgrammingError(msg)
         lfid = url_[len(_lfscheme) :]
         return storefactory.getlfile(ui, lfid)
     else:
         return orig(ui, url_, data=data, **kwargs)

mercurial/cmdutil.py

@@ -1,4009 +1,4102 b''
 # cmdutil.py - help for command processing in mercurial
 #
 # Copyright 2005-2007 Olivia Mackall <[e-mail]>
 #
 # This software may be used and distributed according to the terms of the
 # GNU General Public License version 2 or any later version.
 
 
 import copy as copymod
 import errno
 import os
 import re
 
 from .i18n import _
 from .node import (
     hex,
     nullrev,
     short,
 )
 from .pycompat import (
     getattr,
     open,
     setattr,
 )
 from .thirdparty import attr
 
 from . import (
     bookmarks,
     changelog,
     copies,
     crecord as crecordmod,
-    dirstateguard,
     encoding,
     error,
     formatter,
     logcmdutil,
     match as matchmod,
     merge as mergemod,
     mergestate as mergestatemod,
     mergeutil,
     obsolete,
     patch,
     pathutil,
     phases,
     pycompat,
     repair,
     revlog,
     rewriteutil,
     scmutil,
     state as statemod,
     subrepoutil,
     templatekw,
     templater,
     util,
     vfs as vfsmod,
 )
 
 from .utils import (
     dateutil,
     stringutil,
 )
 
 from .revlogutils import (
     constants as revlog_constants,
 )
 
 if pycompat.TYPE_CHECKING:
     from typing import (
         Any,
         Dict,
     )
 
     for t in (Any, Dict):
         assert t
 
 stringio = util.stringio
 
 # templates of common command options
 
 dryrunopts = [
     (b'n', b'dry-run', None, _(b'do not perform actions, just print output')),
 ]
 
 confirmopts = [
     (b'', b'confirm', None, _(b'ask before applying actions')),
 ]
 
 remoteopts = [
     (b'e', b'ssh', b'', _(b'specify ssh command to use'), _(b'CMD')),
     (
         b'',
         b'remotecmd',
         b'',
         _(b'specify hg command to run on the remote side'),
         _(b'CMD'),
     ),
     (
         b'',
         b'insecure',
         None,
         _(b'do not verify server certificate (ignoring web.cacerts config)'),
     ),
 ]
 
 walkopts = [
     (
         b'I',
         b'include',
         [],
         _(b'include names matching the given patterns'),
         _(b'PATTERN'),
     ),
     (
         b'X',
         b'exclude',
         [],
         _(b'exclude names matching the given patterns'),
         _(b'PATTERN'),
     ),
 ]
 
 commitopts = [
     (b'm', b'message', b'', _(b'use text as commit message'), _(b'TEXT')),
     (b'l', b'logfile', b'', _(b'read commit message from file'), _(b'FILE')),
 ]
 
 commitopts2 = [
     (
         b'd',
         b'date',
         b'',
         _(b'record the specified date as commit date'),
         _(b'DATE'),
     ),
     (
         b'u',
         b'user',
         b'',
         _(b'record the specified user as committer'),
         _(b'USER'),
     ),
 ]
 
 commitopts3 = [
     (b'D', b'currentdate', None, _(b'record the current date as commit date')),
     (b'U', b'currentuser', None, _(b'record the current user as committer')),
 ]
 
 formatteropts = [
     (b'T', b'template', b'', _(b'display with template'), _(b'TEMPLATE')),
 ]
 
 templateopts = [
     (
         b'',
         b'style',
         b'',
         _(b'display using template map file (DEPRECATED)'),
         _(b'STYLE'),
     ),
     (b'T', b'template', b'', _(b'display with template'), _(b'TEMPLATE')),
 ]
 
 logopts = [
     (b'p', b'patch', None, _(b'show patch')),
     (b'g', b'git', None, _(b'use git extended diff format')),
     (b'l', b'limit', b'', _(b'limit number of changes displayed'), _(b'NUM')),
     (b'M', b'no-merges', None, _(b'do not show merges')),
     (b'', b'stat', None, _(b'output diffstat-style summary of changes')),
     (b'G', b'graph', None, _(b"show the revision DAG")),
 ] + templateopts
 
 diffopts = [
     (b'a', b'text', None, _(b'treat all files as text')),
     (
         b'g',
         b'git',
         None,
         _(b'use git extended diff format (DEFAULT: diff.git)'),
     ),
     (b'', b'binary', None, _(b'generate binary diffs in git mode (default)')),
     (b'', b'nodates', None, _(b'omit dates from diff headers')),
 ]
 
 diffwsopts = [
     (
         b'w',
         b'ignore-all-space',
         None,
         _(b'ignore white space when comparing lines'),
     ),
     (
         b'b',
         b'ignore-space-change',
         None,
         _(b'ignore changes in the amount of white space'),
     ),
     (
         b'B',
         b'ignore-blank-lines',
         None,
         _(b'ignore changes whose lines are all blank'),
     ),
     (
         b'Z',
         b'ignore-space-at-eol',
         None,
         _(b'ignore changes in whitespace at EOL'),
     ),
 ]
 
 diffopts2 = (
     [
         (b'', b'noprefix', None, _(b'omit a/ and b/ prefixes from filenames')),
         (
             b'p',
             b'show-function',
             None,
             _(
                 b'show which function each change is in (DEFAULT: diff.showfunc)'
             ),
         ),
         (b'', b'reverse', None, _(b'produce a diff that undoes the changes')),
     ]
     + diffwsopts
     + [
         (
             b'U',
             b'unified',
             b'',
             _(b'number of lines of context to show'),
             _(b'NUM'),
         ),
         (b'', b'stat', None, _(b'output diffstat-style summary of changes')),
         (
             b'',
             b'root',
             b'',
             _(b'produce diffs relative to subdirectory'),
             _(b'DIR'),
         ),
     ]
 )
 
 mergetoolopts = [
     (b't', b'tool', b'', _(b'specify merge tool'), _(b'TOOL')),
 ]
 
 similarityopts = [
     (
         b's',
         b'similarity',
         b'',
         _(b'guess renamed files by similarity (0<=s<=100)'),
         _(b'SIMILARITY'),
     )
 ]
 
 subrepoopts = [(b'S', b'subrepos', None, _(b'recurse into subrepositories'))]
 
 debugrevlogopts = [
     (b'c', b'changelog', False, _(b'open changelog')),
     (b'm', b'manifest', False, _(b'open manifest')),
     (b'', b'dir', b'', _(b'open directory manifest')),
 ]
 
 # special string such that everything below this line will be ingored in the
 # editor text
 _linebelow = b"^HG: ------------------------ >8 ------------------------$"
 
 
 def check_at_most_one_arg(opts, *args):
     """abort if more than one of the arguments are in opts
 
     Returns the unique argument or None if none of them were specified.
     """
 
     def to_display(name):
         return pycompat.sysbytes(name).replace(b'_', b'-')
 
     previous = None
     for x in args:
         if opts.get(x):
             if previous:
                 raise error.InputError(
                     _(b'cannot specify both --%s and --%s')
                     % (to_display(previous), to_display(x))
                 )
             previous = x
     return previous
 
 
 def check_incompatible_arguments(opts, first, others):
     """abort if the first argument is given along with any of the others
 
     Unlike check_at_most_one_arg(), `others` are not mutually exclusive
     among themselves, and they're passed as a single collection.
     """
     for other in others:
         check_at_most_one_arg(opts, first, other)
 
 
 def resolve_commit_options(ui, opts):
     """modify commit options dict to handle related options
 
     The return value indicates that ``rewrite.update-timestamp`` is the reason
     the ``date`` option is set.
     """
     check_at_most_one_arg(opts, 'date', 'currentdate')
     check_at_most_one_arg(opts, 'user', 'currentuser')
 
     datemaydiffer = False  # date-only change should be ignored?
 
     if opts.get('currentdate'):
         opts['date'] = b'%d %d' % dateutil.makedate()
     elif (
         not opts.get('date')
         and ui.configbool(b'rewrite', b'update-timestamp')
         and opts.get('currentdate') is None
     ):
         opts['date'] = b'%d %d' % dateutil.makedate()
         datemaydiffer = True
 
     if opts.get('currentuser'):
         opts['user'] = ui.username()
 
     return datemaydiffer
 
 
 def check_note_size(opts):
     """make sure note is of valid format"""
 
     note = opts.get('note')
     if not note:
         return
 
     if len(note) > 255:
         raise error.InputError(_(b"cannot store a note of more than 255 bytes"))
     if b'\n' in note:
         raise error.InputError(_(b"note cannot contain a newline"))
 
 
 def ishunk(x):
     hunkclasses = (crecordmod.uihunk, patch.recordhunk)
     return isinstance(x, hunkclasses)
 
 
 def isheader(x):
     headerclasses = (crecordmod.uiheader, patch.header)
     return isinstance(x, headerclasses)
 
 
 def newandmodified(chunks):
     newlyaddedandmodifiedfiles = set()
     alsorestore = set()
     for chunk in chunks:
         if isheader(chunk) and chunk.isnewfile():
             newlyaddedandmodifiedfiles.add(chunk.filename())
             alsorestore.update(set(chunk.files()) - {chunk.filename()})
     return newlyaddedandmodifiedfiles, alsorestore
 
 
 def parsealiases(cmd):
     base_aliases = cmd.split(b"|")
     all_aliases = set(base_aliases)
     extra_aliases = []
     for alias in base_aliases:
         if b'-' in alias:
             folded_alias = alias.replace(b'-', b'')
             if folded_alias not in all_aliases:
                 all_aliases.add(folded_alias)
                 extra_aliases.append(folded_alias)
     base_aliases.extend(extra_aliases)
     return base_aliases
 
 
 def setupwrapcolorwrite(ui):
     # wrap ui.write so diff output can be labeled/colorized
     def wrapwrite(orig, *args, **kw):
         label = kw.pop('label', b'')
         for chunk, l in patch.difflabel(lambda: args):
             orig(chunk, label=label + l)
 
     oldwrite = ui.write
 
     def wrap(*args, **kwargs):
         return wrapwrite(oldwrite, *args, **kwargs)
 
     setattr(ui, 'write', wrap)
     return oldwrite
 
 
 def filterchunks(ui, originalhunks, usecurses, testfile, match, operation=None):
     try:
         if usecurses:
             if testfile:
                 recordfn = crecordmod.testdecorator(
                     testfile, crecordmod.testchunkselector
                 )
             else:
                 recordfn = crecordmod.chunkselector
 
             return crecordmod.filterpatch(
                 ui, originalhunks, recordfn, operation
             )
     except crecordmod.fallbackerror as e:
         ui.warn(b'%s\n' % e)
         ui.warn(_(b'falling back to text mode\n'))
 
     return patch.filterpatch(ui, originalhunks, match, operation)
 
 
 def recordfilter(ui, originalhunks, match, operation=None):
     """Prompts the user to filter the originalhunks and return a list of
     selected hunks.
     *operation* is used for to build ui messages to indicate the user what
     kind of filtering they are doing: reverting, committing, shelving, etc.
     (see patch.filterpatch).
     """
     usecurses = crecordmod.checkcurses(ui)
     testfile = ui.config(b'experimental', b'crecordtest')
     oldwrite = setupwrapcolorwrite(ui)
     try:
         newchunks, newopts = filterchunks(
             ui, originalhunks, usecurses, testfile, match, operation
         )
     finally:
         ui.write = oldwrite
     return newchunks, newopts
 
 
 def dorecord(
     ui, repo, commitfunc, cmdsuggest, backupall, filterfn, *pats, **opts
 ):
     opts = pycompat.byteskwargs(opts)
     if not ui.interactive():
         if cmdsuggest:
             msg = _(b'running non-interactively, use %s instead') % cmdsuggest
         else:
             msg = _(b'running non-interactively')
         raise error.InputError(msg)
 
     # make sure username is set before going interactive
     if not opts.get(b'user'):
         ui.username()  # raise exception, username not provided
 
     def recordfunc(ui, repo, message, match, opts):
         """This is generic record driver.
 
         Its job is to interactively filter local changes, and
         accordingly prepare working directory into a state in which the
         job can be delegated to a non-interactive commit command such as
         'commit' or 'qrefresh'.
 
         After the actual job is done by non-interactive command, the
         working directory is restored to its original state.
 
         In the end we'll record interesting changes, and everything else
         will be left in place, so the user can continue working.
         """
         if not opts.get(b'interactive-unshelve'):
             checkunfinished(repo, commit=True)
         wctx = repo[None]
         merge = len(wctx.parents()) > 1
         if merge:
             raise error.InputError(
                 _(
                     b'cannot partially commit a merge '
                     b'(use "hg commit" instead)'
                 )
             )
 
         def fail(f, msg):
             raise error.InputError(b'%s: %s' % (f, msg))
 
         force = opts.get(b'force')
         if not force:
             match = matchmod.badmatch(match, fail)
 
         status = repo.status(match=match)
 
         overrides = {(b'ui', b'commitsubrepos'): True}
 
         with repo.ui.configoverride(overrides, b'record'):
             # subrepoutil.precommit() modifies the status
             tmpstatus = scmutil.status(
                 copymod.copy(status.modified),
                 copymod.copy(status.added),
                 copymod.copy(status.removed),
                 copymod.copy(status.deleted),
                 copymod.copy(status.unknown),
                 copymod.copy(status.ignored),
                 copymod.copy(status.clean),  # pytype: disable=wrong-arg-count
             )
 
             # Force allows -X subrepo to skip the subrepo.
             subs, commitsubs, newstate = subrepoutil.precommit(
                 repo.ui, wctx, tmpstatus, match, force=True
             )
             for s in subs:
                 if s in commitsubs:
                     dirtyreason = wctx.sub(s).dirtyreason(True)
                     raise error.Abort(dirtyreason)
 
         if not force:
             repo.checkcommitpatterns(wctx, match, status, fail)
         diffopts = patch.difffeatureopts(
             ui,
             opts=opts,
             whitespace=True,
             section=b'commands',
             configprefix=b'commit.interactive.',
         )
         diffopts.nodates = True
         diffopts.git = True
         diffopts.showfunc = True
         originaldiff = patch.diff(repo, changes=status, opts=diffopts)
         original_headers = patch.parsepatch(originaldiff)
         match = scmutil.match(repo[None], pats)
 
         # 1. filter patch, since we are intending to apply subset of it
         try:
             chunks, newopts = filterfn(ui, original_headers, match)
         except error.PatchParseError as err:
             raise error.InputError(_(b'error parsing patch: %s') % err)
         except error.PatchApplicationError as err:
             raise error.StateError(_(b'error applying patch: %s') % err)
         opts.update(newopts)
 
         # We need to keep a backup of files that have been newly added and
         # modified during the recording process because there is a previous
         # version without the edit in the workdir. We also will need to restore
         # files that were the sources of renames so that the patch application
         # works.
         newlyaddedandmodifiedfiles, alsorestore = newandmodified(chunks)
         contenders = set()
         for h in chunks:
             if isheader(h):
                 contenders.update(set(h.files()))
 
         changed = status.modified + status.added + status.removed
         newfiles = [f for f in changed if f in contenders]
         if not newfiles:
             ui.status(_(b'no changes to record\n'))
             return 0
 
         modified = set(status.modified)
 
         # 2. backup changed files, so we can restore them in the end
 
         if backupall:
             tobackup = changed
         else:
             tobackup = [
                 f
                 for f in newfiles
                 if f in modified or f in newlyaddedandmodifiedfiles
             ]
         backups = {}
         if tobackup:
             backupdir = repo.vfs.join(b'record-backups')
             try:
                 os.mkdir(backupdir)
             except FileExistsError:
                 pass
         try:
             # backup continues
             for f in tobackup:
                 fd, tmpname = pycompat.mkstemp(
                     prefix=os.path.basename(f) + b'.', dir=backupdir
                 )
                 os.close(fd)
                 ui.debug(b'backup %r as %r\n' % (f, tmpname))
                 util.copyfile(repo.wjoin(f), tmpname, copystat=True)
                 backups[f] = tmpname
 
             fp = stringio()
             for c in chunks:
                 fname = c.filename()
                 if fname in backups:
                     c.write(fp)
             dopatch = fp.tell()
             fp.seek(0)
 
             # 2.5 optionally review / modify patch in text editor
             if opts.get(b'review', False):
                 patchtext = (
                     crecordmod.diffhelptext
                     + crecordmod.patchhelptext
                     + fp.read()
                 )
                 reviewedpatch = ui.edit(
                     patchtext, b"", action=b"diff", repopath=repo.path
                 )
                 fp.truncate(0)
                 fp.write(reviewedpatch)
                 fp.seek(0)
 
             [os.unlink(repo.wjoin(c)) for c in newlyaddedandmodifiedfiles]
             # 3a. apply filtered patch to clean repo  (clean)
             if backups:
                 m = scmutil.matchfiles(repo, set(backups.keys()) | alsorestore)
                 mergemod.revert_to(repo[b'.'], matcher=m)
 
             # 3b. (apply)
             if dopatch:
                 try:
                     ui.debug(b'applying patch\n')
                     ui.debug(fp.getvalue())
                     patch.internalpatch(ui, repo, fp, 1, eolmode=None)
                 except error.PatchParseError as err:
                     raise error.InputError(pycompat.bytestr(err))
                 except error.PatchApplicationError as err:
                     raise error.StateError(pycompat.bytestr(err))
             del fp
 
             # 4. We prepared working directory according to filtered
             #    patch. Now is the time to delegate the job to
             #    commit/qrefresh or the like!
 
             # Make all of the pathnames absolute.
             newfiles = [repo.wjoin(nf) for nf in newfiles]
             return commitfunc(ui, repo, *newfiles, **pycompat.strkwargs(opts))
         finally:
             # 5. finally restore backed-up files
             try:
                 dirstate = repo.dirstate
                 for realname, tmpname in backups.items():
                     ui.debug(b'restoring %r to %r\n' % (tmpname, realname))
 
                     if dirstate.get_entry(realname).maybe_clean:
                         # without normallookup, restoring timestamp
                         # may cause partially committed files
                         # to be treated as unmodified
 
                         # XXX-PENDINGCHANGE: We should clarify the context in
                         # which this function is called  to make sure it
                         # already called within a `pendingchange`, However we
                         # are taking a shortcut here in order to be able to
                         # quickly deprecated the older API.
                         with dirstate.changing_parents(repo):
                             dirstate.update_file(
                                 realname,
                                 p1_tracked=True,
                                 wc_tracked=True,
                                 possibly_dirty=True,
                             )
 
                     # copystat=True here and above are a hack to trick any
                     # editors that have f open that we haven't modified them.
                     #
                     # Also note that this racy as an editor could notice the
                     # file's mtime before we've finished writing it.
                     util.copyfile(tmpname, repo.wjoin(realname), copystat=True)
                     os.unlink(tmpname)
                 if tobackup:
                     os.rmdir(backupdir)
             except OSError:
                 pass
 
     def recordinwlock(ui, repo, message, match, opts):
         with repo.wlock():
             return recordfunc(ui, repo, message, match, opts)
 
     return commit(ui, repo, recordinwlock, pats, opts)
 
 
 class dirnode:
     """
     Represent a directory in user working copy with information required for
     the purpose of tersing its status.
 
     path is the path to the directory, without a trailing '/'
 
     statuses is a set of statuses of all files in this directory (this includes
     all the files in all the subdirectories too)
 
     files is a list of files which are direct child of this directory
 
     subdirs is a dictionary of sub-directory name as the key and it's own
     dirnode object as the value
     """
 
     def __init__(self, dirpath):
         self.path = dirpath
         self.statuses = set()
         self.files = []
         self.subdirs = {}
 
     def _addfileindir(self, filename, status):
         """Add a file in this directory as a direct child."""
         self.files.append((filename, status))
 
     def addfile(self, filename, status):
         """
         Add a file to this directory or to its direct parent directory.
 
         If the file is not direct child of this directory, we traverse to the
         directory of which this file is a direct child of and add the file
         there.
         """
 
         # the filename contains a path separator, it means it's not the direct
         # child of this directory
         if b'/' in filename:
             subdir, filep = filename.split(b'/', 1)
 
             # does the dirnode object for subdir exists
             if subdir not in self.subdirs:
                 subdirpath = pathutil.join(self.path, subdir)
                 self.subdirs[subdir] = dirnode(subdirpath)
 
             # try adding the file in subdir
             self.subdirs[subdir].addfile(filep, status)
 
         else:
             self._addfileindir(filename, status)
 
         if status not in self.statuses:
             self.statuses.add(status)
 
     def iterfilepaths(self):
         """Yield (status, path) for files directly under this directory."""
         for f, st in self.files:
             yield st, pathutil.join(self.path, f)
 
     def tersewalk(self, terseargs):
         """
         Yield (status, path) obtained by processing the status of this
         dirnode.
 
         terseargs is the string of arguments passed by the user with `--terse`
         flag.
 
         Following are the cases which can happen:
 
         1) All the files in the directory (including all the files in its
         subdirectories) share the same status and the user has asked us to terse
         that status. -> yield (status, dirpath).  dirpath will end in '/'.
 
         2) Otherwise, we do following:
 
                 a) Yield (status, filepath)  for all the files which are in this
                     directory (only the ones in this directory, not the subdirs)
 
                 b) Recurse the function on all the subdirectories of this
                    directory
         """
 
         if len(self.statuses) == 1:
             onlyst = self.statuses.pop()
 
             # Making sure we terse only when the status abbreviation is
             # passed as terse argument
             if onlyst in terseargs:
                 yield onlyst, self.path + b'/'
                 return
 
         # add the files to status list
         for st, fpath in self.iterfilepaths():
             yield st, fpath
 
         # recurse on the subdirs
         for dirobj in self.subdirs.values():
             for st, fpath in dirobj.tersewalk(terseargs):
                 yield st, fpath
 
 
 def tersedir(statuslist, terseargs):
     """
     Terse the status if all the files in a directory shares the same status.
 
     statuslist is scmutil.status() object which contains a list of files for
     each status.
     terseargs is string which is passed by the user as the argument to `--terse`
     flag.
 
     The function makes a tree of objects of dirnode class, and at each node it
     stores the information required to know whether we can terse a certain
     directory or not.
     """
     # the order matters here as that is used to produce final list
     allst = (b'm', b'a', b'r', b'd', b'u', b'i', b'c')
 
     # checking the argument validity
     for s in pycompat.bytestr(terseargs):
         if s not in allst:
             raise error.InputError(_(b"'%s' not recognized") % s)
 
     # creating a dirnode object for the root of the repo
     rootobj = dirnode(b'')
     pstatus = (
         b'modified',
         b'added',
         b'deleted',
         b'clean',
         b'unknown',
         b'ignored',
         b'removed',
     )
 
     tersedict = {}
     for attrname in pstatus:
         statuschar = attrname[0:1]
         for f in getattr(statuslist, attrname):
             rootobj.addfile(f, statuschar)
         tersedict[statuschar] = []
 
     # we won't be tersing the root dir, so add files in it
     for st, fpath in rootobj.iterfilepaths():
         tersedict[st].append(fpath)
 
     # process each sub-directory and build tersedict
     for subdir in rootobj.subdirs.values():
         for st, f in subdir.tersewalk(terseargs):
             tersedict[st].append(f)
 
     tersedlist = []
     for st in allst:
         tersedict[st].sort()
         tersedlist.append(tersedict[st])
 
     return scmutil.status(*tersedlist)
 
 
 def _commentlines(raw):
     '''Surround lineswith a comment char and a new line'''
     lines = raw.splitlines()
     commentedlines = [b'# %s' % line for line in lines]
     return b'\n'.join(commentedlines) + b'\n'
 
 
 @attr.s(frozen=True)
 class morestatus:
     repo = attr.ib()
     unfinishedop = attr.ib()
     unfinishedmsg = attr.ib()
     activemerge = attr.ib()
     unresolvedpaths = attr.ib()
     _formattedpaths = attr.ib(init=False, default=set())
     _label = b'status.morestatus'
 
     def formatfile(self, path, fm):
         self._formattedpaths.add(path)
         if self.activemerge and path in self.unresolvedpaths:
             fm.data(unresolved=True)
 
     def formatfooter(self, fm):
         if self.unfinishedop or self.unfinishedmsg:
             fm.startitem()
             fm.data(itemtype=b'morestatus')
 
         if self.unfinishedop:
             fm.data(unfinished=self.unfinishedop)
             statemsg = (
                 _(b'The repository is in an unfinished *%s* state.')
                 % self.unfinishedop
             )
             fm.plain(b'%s\n' % _commentlines(statemsg), label=self._label)
         if self.unfinishedmsg:
             fm.data(unfinishedmsg=self.unfinishedmsg)
 
         # May also start new data items.
         self._formatconflicts(fm)
 
         if self.unfinishedmsg:
             fm.plain(
                 b'%s\n' % _commentlines(self.unfinishedmsg), label=self._label
             )
 
     def _formatconflicts(self, fm):
         if not self.activemerge:
             return
 
         if self.unresolvedpaths:
             mergeliststr = b'\n'.join(
                 [
                     b'    %s'
                     % util.pathto(self.repo.root, encoding.getcwd(), path)
                     for path in self.unresolvedpaths
                 ]
             )
             msg = (
                 _(
                     b'''Unresolved merge conflicts:
 
 %s
 
 To mark files as resolved:  hg resolve --mark FILE'''
                 )
                 % mergeliststr
             )
 
             # If any paths with unresolved conflicts were not previously
             # formatted, output them now.
             for f in self.unresolvedpaths:
                 if f in self._formattedpaths:
                     # Already output.
                     continue
                 fm.startitem()
                 fm.context(repo=self.repo)
                 # We can't claim to know the status of the file - it may just
                 # have been in one of the states that were not requested for
                 # display, so it could be anything.
                 fm.data(itemtype=b'file', path=f, unresolved=True)
 
         else:
             msg = _(b'No unresolved merge conflicts.')
 
         fm.plain(b'%s\n' % _commentlines(msg), label=self._label)
 
 
 def readmorestatus(repo):
     """Returns a morestatus object if the repo has unfinished state."""
     statetuple = statemod.getrepostate(repo)
     mergestate = mergestatemod.mergestate.read(repo)
     activemerge = mergestate.active()
     if not statetuple and not activemerge:
         return None
 
     unfinishedop = unfinishedmsg = unresolved = None
     if statetuple:
         unfinishedop, unfinishedmsg = statetuple
     if activemerge:
         unresolved = sorted(mergestate.unresolved())
     return morestatus(
         repo, unfinishedop, unfinishedmsg, activemerge, unresolved
     )
 
 
 def findpossible(cmd, table, strict=False):
     """
     Return cmd -> (aliases, command table entry)
     for each matching command.
     Return debug commands (or their aliases) only if no normal command matches.
     """
     choice = {}
     debugchoice = {}
 
     if cmd in table:
         # short-circuit exact matches, "log" alias beats "log|history"
         keys = [cmd]
     else:
         keys = table.keys()
 
     allcmds = []
     for e in keys:
         aliases = parsealiases(e)
         allcmds.extend(aliases)
         found = None
         if cmd in aliases:
             found = cmd
         elif not strict:
             for a in aliases:
                 if a.startswith(cmd):
                     found = a
                     break
         if found is not None:
             if aliases[0].startswith(b"debug") or found.startswith(b"debug"):
                 debugchoice[found] = (aliases, table[e])
             else:
                 choice[found] = (aliases, table[e])
 
     if not choice and debugchoice:
         choice = debugchoice
 
     return choice, allcmds
 
 
 def findcmd(cmd, table, strict=True):
     """Return (aliases, command table entry) for command string."""
     choice, allcmds = findpossible(cmd, table, strict)
 
     if cmd in choice:
         return choice[cmd]
 
     if len(choice) > 1:
         clist = sorted(choice)
         raise error.AmbiguousCommand(cmd, clist)
 
     if choice:
         return list(choice.values())[0]
 
     raise error.UnknownCommand(cmd, allcmds)
 
 
 def changebranch(ui, repo, revs, label, opts):
     """Change the branch name of given revs to label"""
 
     with repo.wlock(), repo.lock(), repo.transaction(b'branches'):
         # abort in case of uncommitted merge or dirty wdir
         bailifchanged(repo)
         revs = logcmdutil.revrange(repo, revs)
         if not revs:
             raise error.InputError(b"empty revision set")
         roots = repo.revs(b'roots(%ld)', revs)
         if len(roots) > 1:
             raise error.InputError(
                 _(b"cannot change branch of non-linear revisions")
             )
         rewriteutil.precheck(repo, revs, b'change branch of')
 
         root = repo[roots.first()]
         rpb = {parent.branch() for parent in root.parents()}
         if (
             not opts.get(b'force')
             and label not in rpb
             and label in repo.branchmap()
         ):
             raise error.InputError(
                 _(b"a branch of the same name already exists")
             )
 
         # make sure only topological heads
         if repo.revs(b'heads(%ld) - head()', revs):
             raise error.InputError(
                 _(b"cannot change branch in middle of a stack")
             )
 
         replacements = {}
         # avoid import cycle mercurial.cmdutil -> mercurial.context ->
         # mercurial.subrepo -> mercurial.cmdutil
         from . import context
 
         for rev in revs:
             ctx = repo[rev]
             oldbranch = ctx.branch()
             # check if ctx has same branch
             if oldbranch == label:
                 continue
 
             def filectxfn(repo, newctx, path):
                 try:
                     return ctx[path]
                 except error.ManifestLookupError:
                     return None
 
             ui.debug(
                 b"changing branch of '%s' from '%s' to '%s'\n"
                 % (hex(ctx.node()), oldbranch, label)
             )
             extra = ctx.extra()
             extra[b'branch_change'] = hex(ctx.node())
             # While changing branch of set of linear commits, make sure that
             # we base our commits on new parent rather than old parent which
             # was obsoleted while changing the branch
             p1 = ctx.p1().node()
             p2 = ctx.p2().node()
             if p1 in replacements:
                 p1 = replacements[p1][0]
             if p2 in replacements:
                 p2 = replacements[p2][0]
 
             mc = context.memctx(
                 repo,
                 (p1, p2),
                 ctx.description(),
                 ctx.files(),
                 filectxfn,
                 user=ctx.user(),
                 date=ctx.date(),
                 extra=extra,
                 branch=label,
             )
 
             newnode = repo.commitctx(mc)
             replacements[ctx.node()] = (newnode,)
             ui.debug(b'new node id is %s\n' % hex(newnode))
 
         # create obsmarkers and move bookmarks
         scmutil.cleanupnodes(
             repo, replacements, b'branch-change', fixphase=True
         )
 
         # move the working copy too
         wctx = repo[None]
         # in-progress merge is a bit too complex for now.
         if len(wctx.parents()) == 1:
             newid = replacements.get(wctx.p1().node())
             if newid is not None:
                 # avoid import cycle mercurial.cmdutil -> mercurial.hg ->
                 # mercurial.cmdutil
                 from . import hg
 
                 hg.update(repo, newid[0], quietempty=True)
 
         ui.status(_(b"changed branch on %d changesets\n") % len(replacements))
 
 
 def findrepo(p):
     while not os.path.isdir(os.path.join(p, b".hg")):
         oldp, p = p, os.path.dirname(p)
         if p == oldp:
             return None
 
     return p
 
 
 def bailifchanged(repo, merge=True, hint=None):
     """enforce the precondition that working directory must be clean.
 
     'merge' can be set to false if a pending uncommitted merge should be
     ignored (such as when 'update --check' runs).
 
     'hint' is the usual hint given to Abort exception.
     """
 
     if merge and repo.dirstate.p2() != repo.nullid:
         raise error.StateError(_(b'outstanding uncommitted merge'), hint=hint)
     st = repo.status()
     if st.modified or st.added or st.removed or st.deleted:
         raise error.StateError(_(b'uncommitted changes'), hint=hint)
     ctx = repo[None]
     for s in sorted(ctx.substate):
         ctx.sub(s).bailifchanged(hint=hint)
 
 
 def logmessage(ui, opts):
     """get the log message according to -m and -l option"""
 
     check_at_most_one_arg(opts, b'message', b'logfile')
 
     message = opts.get(b'message')
     logfile = opts.get(b'logfile')
 
     if not message and logfile:
         try:
             if isstdiofilename(logfile):
                 message = ui.fin.read()
             else:
                 message = b'\n'.join(util.readfile(logfile).splitlines())
         except IOError as inst:
             raise error.Abort(
                 _(b"can't read commit message '%s': %s")
                 % (logfile, encoding.strtolocal(inst.strerror))
             )
     return message
 
 
 def mergeeditform(ctxorbool, baseformname):
     """return appropriate editform name (referencing a committemplate)
 
     'ctxorbool' is either a ctx to be committed, or a bool indicating whether
     merging is committed.
 
     This returns baseformname with '.merge' appended if it is a merge,
     otherwise '.normal' is appended.
     """
     if isinstance(ctxorbool, bool):
         if ctxorbool:
             return baseformname + b".merge"
     elif len(ctxorbool.parents()) > 1:
         return baseformname + b".merge"
 
     return baseformname + b".normal"
 
 
 def getcommiteditor(
     edit=False, finishdesc=None, extramsg=None, editform=b'', **opts
 ):
     """get appropriate commit message editor according to '--edit' option
 
     'finishdesc' is a function to be called with edited commit message
     (= 'description' of the new changeset) just after editing, but
     before checking empty-ness. It should return actual text to be
     stored into history. This allows to change description before
     storing.
 
     'extramsg' is a extra message to be shown in the editor instead of
     'Leave message empty to abort commit' line. 'HG: ' prefix and EOL
     is automatically added.
 
     'editform' is a dot-separated list of names, to distinguish
     the purpose of commit text editing.
 
     'getcommiteditor' returns 'commitforceeditor' regardless of
     'edit', if one of 'finishdesc' or 'extramsg' is specified, because
     they are specific for usage in MQ.
     """
     if edit or finishdesc or extramsg:
         return lambda r, c, s: commitforceeditor(
             r, c, s, finishdesc=finishdesc, extramsg=extramsg, editform=editform
         )
     elif editform:
         return lambda r, c, s: commiteditor(r, c, s, editform=editform)
     else:
         return commiteditor
 
 
 def _escapecommandtemplate(tmpl):
     parts = []
     for typ, start, end in templater.scantemplate(tmpl, raw=True):
         if typ == b'string':
             parts.append(stringutil.escapestr(tmpl[start:end]))
         else:
             parts.append(tmpl[start:end])
     return b''.join(parts)
 
 
 def rendercommandtemplate(ui, tmpl, props):
     r"""Expand a literal template 'tmpl' in a way suitable for command line
 
     '\' in outermost string is not taken as an escape character because it
     is a directory separator on Windows.
 
     >>> from . import ui as uimod
     >>> ui = uimod.ui()
     >>> rendercommandtemplate(ui, b'c:\\{path}', {b'path': b'foo'})
     'c:\\foo'
     >>> rendercommandtemplate(ui, b'{"c:\\{path}"}', {'path': b'foo'})
     'c:{path}'
     """
     if not tmpl:
         return tmpl
     t = formatter.maketemplater(ui, _escapecommandtemplate(tmpl))
     return t.renderdefault(props)
 
 
 def rendertemplate(ctx, tmpl, props=None):
     """Expand a literal template 'tmpl' byte-string against one changeset
 
     Each props item must be a stringify-able value or a callable returning
     such value, i.e. no bare list nor dict should be passed.
     """
     repo = ctx.repo()
     tres = formatter.templateresources(repo.ui, repo)
     t = formatter.maketemplater(
         repo.ui, tmpl, defaults=templatekw.keywords, resources=tres
     )
     mapping = {b'ctx': ctx}
     if props:
         mapping.update(props)
     return t.renderdefault(mapping)
 
 
 def format_changeset_summary(ui, ctx, command=None, default_spec=None):
     """Format a changeset summary (one line)."""
     spec = None
     if command:
         spec = ui.config(
             b'command-templates', b'oneline-summary.%s' % command, None
         )
     if not spec:
         spec = ui.config(b'command-templates', b'oneline-summary')
     if not spec:
         spec = default_spec
     if not spec:
         spec = (
             b'{separate(" ", '
             b'label("oneline-summary.changeset", "{rev}:{node|short}")'
             b', '
             b'join(filter(namespaces % "{ifeq(namespace, "branches", "", join(names % "{label("oneline-summary.{namespace}", name)}", " "))}"), " ")'
             b')} '
             b'"{label("oneline-summary.desc", desc|firstline)}"'
         )
     text = rendertemplate(ctx, spec)
     return text.split(b'\n')[0]
 
 
 def _buildfntemplate(pat, total=None, seqno=None, revwidth=None, pathname=None):
     r"""Convert old-style filename format string to template string
 
     >>> _buildfntemplate(b'foo-%b-%n.patch', seqno=0)
     'foo-{reporoot|basename}-{seqno}.patch'
     >>> _buildfntemplate(b'%R{tags % "{tag}"}%H')
     '{rev}{tags % "{tag}"}{node}'
 
     '\' in outermost strings has to be escaped because it is a directory
     separator on Windows:
 
     >>> _buildfntemplate(b'c:\\tmp\\%R\\%n.patch', seqno=0)
     'c:\\\\tmp\\\\{rev}\\\\{seqno}.patch'
     >>> _buildfntemplate(b'\\\\foo\\bar.patch')
     '\\\\\\\\foo\\\\bar.patch'
     >>> _buildfntemplate(b'\\{tags % "{tag}"}')
     '\\\\{tags % "{tag}"}'
 
     but inner strings follow the template rules (i.e. '\' is taken as an
     escape character):
 
     >>> _buildfntemplate(br'{"c:\tmp"}', seqno=0)
     '{"c:\\tmp"}'
     """
     expander = {
         b'H': b'{node}',
         b'R': b'{rev}',
         b'h': b'{node|short}',
         b'm': br'{sub(r"[^\w]", "_", desc|firstline)}',
         b'r': b'{if(revwidth, pad(rev, revwidth, "0", left=True), rev)}',
         b'%': b'%',
         b'b': b'{reporoot|basename}',
     }
     if total is not None:
         expander[b'N'] = b'{total}'
     if seqno is not None:
         expander[b'n'] = b'{seqno}'
     if total is not None and seqno is not None:
         expander[b'n'] = b'{pad(seqno, total|stringify|count, "0", left=True)}'
     if pathname is not None:
         expander[b's'] = b'{pathname|basename}'
         expander[b'd'] = b'{if(pathname|dirname, pathname|dirname, ".")}'
         expander[b'p'] = b'{pathname}'
 
     newname = []
     for typ, start, end in templater.scantemplate(pat, raw=True):
         if typ != b'string':
             newname.append(pat[start:end])
             continue
         i = start
         while i < end:
             n = pat.find(b'%', i, end)
             if n < 0:
                 newname.append(stringutil.escapestr(pat[i:end]))
                 break
             newname.append(stringutil.escapestr(pat[i:n]))
             if n + 2 > end:
                 raise error.Abort(
                     _(b"incomplete format spec in output filename")
                 )
             c = pat[n + 1 : n + 2]
             i = n + 2
             try:
                 newname.append(expander[c])
             except KeyError:
                 raise error.Abort(
                     _(b"invalid format spec '%%%s' in output filename") % c
                 )
     return b''.join(newname)
 
 
 def makefilename(ctx, pat, **props):
     if not pat:
         return pat
     tmpl = _buildfntemplate(pat, **props)
     # BUG: alias expansion shouldn't be made against template fragments
     # rewritten from %-format strings, but we have no easy way to partially
     # disable the expansion.
     return rendertemplate(ctx, tmpl, pycompat.byteskwargs(props))
 
 
 def isstdiofilename(pat):
     """True if the given pat looks like a filename denoting stdin/stdout"""
     return not pat or pat == b'-'
 
 
 class _unclosablefile:
     def __init__(self, fp):
         self._fp = fp
 
     def close(self):
         pass
 
     def __iter__(self):
         return iter(self._fp)
 
     def __getattr__(self, attr):
         return getattr(self._fp, attr)
 
     def __enter__(self):
         return self
 
     def __exit__(self, exc_type, exc_value, exc_tb):
         pass
 
 
 def makefileobj(ctx, pat, mode=b'wb', **props):
     writable = mode not in (b'r', b'rb')
 
     if isstdiofilename(pat):
         repo = ctx.repo()
         if writable:
             fp = repo.ui.fout
         else:
             fp = repo.ui.fin
         return _unclosablefile(fp)
     fn = makefilename(ctx, pat, **props)
     return open(fn, mode)
 
 
 def openstorage(repo, cmd, file_, opts, returnrevlog=False):
     """opens the changelog, manifest, a filelog or a given revlog"""
     cl = opts[b'changelog']
     mf = opts[b'manifest']
     dir = opts[b'dir']
     msg = None
     if cl and mf:
         msg = _(b'cannot specify --changelog and --manifest at the same time')
     elif cl and dir:
         msg = _(b'cannot specify --changelog and --dir at the same time')
     elif cl or mf or dir:
         if file_:
             msg = _(b'cannot specify filename with --changelog or --manifest')
         elif not repo:
             msg = _(
                 b'cannot specify --changelog or --manifest or --dir '
                 b'without a repository'
             )
     if msg:
         raise error.InputError(msg)
 
     r = None
     if repo:
         if cl:
             r = repo.unfiltered().changelog
         elif dir:
             if not scmutil.istreemanifest(repo):
                 raise error.InputError(
                     _(
                         b"--dir can only be used on repos with "
                         b"treemanifest enabled"
                     )
                 )
             if not dir.endswith(b'/'):
                 dir = dir + b'/'
             dirlog = repo.manifestlog.getstorage(dir)
             if len(dirlog):
                 r = dirlog
         elif mf:
             r = repo.manifestlog.getstorage(b'')
         elif file_:
             filelog = repo.file(file_)
             if len(filelog):
                 r = filelog
 
         # Not all storage may be revlogs. If requested, try to return an actual
         # revlog instance.
         if returnrevlog:
             if isinstance(r, revlog.revlog):
                 pass
             elif util.safehasattr(r, b'_revlog'):
                 r = r._revlog  # pytype: disable=attribute-error
             elif r is not None:
                 raise error.InputError(
                     _(b'%r does not appear to be a revlog') % r
                 )
 
     if not r:
         if not returnrevlog:
             raise error.InputError(_(b'cannot give path to non-revlog'))
 
         if not file_:
             raise error.CommandError(cmd, _(b'invalid arguments'))
         if not os.path.isfile(file_):
             raise error.InputError(_(b"revlog '%s' not found") % file_)
 
         target = (revlog_constants.KIND_OTHER, b'free-form:%s' % file_)
         r = revlog.revlog(
             vfsmod.vfs(encoding.getcwd(), audit=False),
             target=target,
             radix=file_[:-2],
         )
     return r
 
 
 def openrevlog(repo, cmd, file_, opts):
     """Obtain a revlog backing storage of an item.
 
     This is similar to ``openstorage()`` except it always returns a revlog.
 
     In most cases, a caller cares about the main storage object - not the
     revlog backing it. Therefore, this function should only be used by code
     that needs to examine low-level revlog implementation details. e.g. debug
     commands.
     """
     return openstorage(repo, cmd, file_, opts, returnrevlog=True)
 
 
 def copy(ui, repo, pats, opts, rename=False):
     check_incompatible_arguments(opts, b'forget', [b'dry_run'])
 
     # called with the repo lock held
     #
     # hgsep => pathname that uses "/" to separate directories
     # ossep => pathname that uses os.sep to separate directories
     cwd = repo.getcwd()
     targets = {}
     forget = opts.get(b"forget")
     after = opts.get(b"after")
     dryrun = opts.get(b"dry_run")
     rev = opts.get(b'at_rev')
     if rev:
         if not forget and not after:
             # TODO: Remove this restriction and make it also create the copy
             #       targets (and remove the rename source if rename==True).
             raise error.InputError(_(b'--at-rev requires --after'))
         ctx = logcmdutil.revsingle(repo, rev)
         if len(ctx.parents()) > 1:
             raise error.InputError(
                 _(b'cannot mark/unmark copy in merge commit')
             )
     else:
         ctx = repo[None]
 
     pctx = ctx.p1()
 
     uipathfn = scmutil.getuipathfn(repo, legacyrelativevalue=True)
 
     if forget:
         if ctx.rev() is None:
             new_ctx = ctx
         else:
             if len(ctx.parents()) > 1:
                 raise error.InputError(_(b'cannot unmark copy in merge commit'))
             # avoid cycle context -> subrepo -> cmdutil
             from . import context
 
             rewriteutil.precheck(repo, [ctx.rev()], b'uncopy')
             new_ctx = context.overlayworkingctx(repo)
             new_ctx.setbase(ctx.p1())
             mergemod.graft(repo, ctx, wctx=new_ctx)
 
         match = scmutil.match(ctx, pats, opts)
 
         current_copies = ctx.p1copies()
         current_copies.update(ctx.p2copies())
 
         uipathfn = scmutil.getuipathfn(repo)
         for f in ctx.walk(match):
             if f in current_copies:
                 new_ctx[f].markcopied(None)
             elif match.exact(f):
                 ui.warn(
                     _(
                         b'%s: not unmarking as copy - file is not marked as copied\n'
                     )
                     % uipathfn(f)
                 )
 
         if ctx.rev() is not None:
             with repo.lock():
                 mem_ctx = new_ctx.tomemctx_for_amend(ctx)
                 new_node = mem_ctx.commit()
 
                 if repo.dirstate.p1() == ctx.node():
                     with repo.dirstate.changing_parents(repo):
                         scmutil.movedirstate(repo, repo[new_node])
                 replacements = {ctx.node(): [new_node]}
                 scmutil.cleanupnodes(
                     repo, replacements, b'uncopy', fixphase=True
                 )
 
         return
 
     pats = scmutil.expandpats(pats)
     if not pats:
         raise error.InputError(_(b'no source or destination specified'))
     if len(pats) == 1:
         raise error.InputError(_(b'no destination specified'))
     dest = pats.pop()
 
     def walkpat(pat):
         srcs = []
         # TODO: Inline and simplify the non-working-copy version of this code
         # since it shares very little with the working-copy version of it.
         ctx_to_walk = ctx if ctx.rev() is None else pctx
         m = scmutil.match(ctx_to_walk, [pat], opts, globbed=True)
         for abs in ctx_to_walk.walk(m):
             rel = uipathfn(abs)
             exact = m.exact(abs)
             if abs not in ctx:
                 if abs in pctx:
                     if not after:
                         if exact:
                             ui.warn(
                                 _(
                                     b'%s: not copying - file has been marked '
                                     b'for remove\n'
                                 )
                                 % rel
                             )
                         continue
                 else:
                     if exact:
                         ui.warn(
                             _(b'%s: not copying - file is not managed\n') % rel
                         )
                     continue
 
             # abs: hgsep
             # rel: ossep
             srcs.append((abs, rel, exact))
         return srcs
 
     if ctx.rev() is not None:
         rewriteutil.precheck(repo, [ctx.rev()], b'uncopy')
         absdest = pathutil.canonpath(repo.root, cwd, dest)
         if ctx.hasdir(absdest):
             raise error.InputError(
                 _(b'%s: --at-rev does not support a directory as destination')
                 % uipathfn(absdest)
             )
         if absdest not in ctx:
             raise error.InputError(
                 _(b'%s: copy destination does not exist in %s')
                 % (uipathfn(absdest), ctx)
             )
 
         # avoid cycle context -> subrepo -> cmdutil
         from . import context
 
         copylist = []
         for pat in pats:
             srcs = walkpat(pat)
             if not srcs:
                 continue
             for abs, rel, exact in srcs:
                 copylist.append(abs)
 
         if not copylist:
             raise error.InputError(_(b'no files to copy'))
         # TODO: Add support for `hg cp --at-rev . foo bar dir` and
         # `hg cp --at-rev . dir1 dir2`, preferably unifying the code with the
         # existing functions below.
         if len(copylist) != 1:
             raise error.InputError(_(b'--at-rev requires a single source'))
 
         new_ctx = context.overlayworkingctx(repo)
         new_ctx.setbase(ctx.p1())
         mergemod.graft(repo, ctx, wctx=new_ctx)
 
         new_ctx.markcopied(absdest, copylist[0])
 
         with repo.lock():
             mem_ctx = new_ctx.tomemctx_for_amend(ctx)
             new_node = mem_ctx.commit()
 
             if repo.dirstate.p1() == ctx.node():
                 with repo.dirstate.changing_parents(repo):
                     scmutil.movedirstate(repo, repo[new_node])
             replacements = {ctx.node(): [new_node]}
             scmutil.cleanupnodes(repo, replacements, b'copy', fixphase=True)
 
         return
 
     # abssrc: hgsep
     # relsrc: ossep
     # otarget: ossep
     def copyfile(abssrc, relsrc, otarget, exact):
         abstarget = pathutil.canonpath(repo.root, cwd, otarget)
         if b'/' in abstarget:
             # We cannot normalize abstarget itself, this would prevent
             # case only renames, like a => A.
             abspath, absname = abstarget.rsplit(b'/', 1)
             abstarget = repo.dirstate.normalize(abspath) + b'/' + absname
         reltarget = repo.pathto(abstarget, cwd)
         target = repo.wjoin(abstarget)
         src = repo.wjoin(abssrc)
         entry = repo.dirstate.get_entry(abstarget)
 
         already_commited = entry.tracked and not entry.added
 
         scmutil.checkportable(ui, abstarget)
 
         # check for collisions
         prevsrc = targets.get(abstarget)
         if prevsrc is not None:
             ui.warn(
                 _(b'%s: not overwriting - %s collides with %s\n')
                 % (
                     reltarget,
                     repo.pathto(abssrc, cwd),
                     repo.pathto(prevsrc, cwd),
                 )
             )
             return True  # report a failure
 
         # check for overwrites
         exists = os.path.lexists(target)
         samefile = False
         if exists and abssrc != abstarget:
             if repo.dirstate.normalize(abssrc) == repo.dirstate.normalize(
                 abstarget
             ):
                 if not rename:
                     ui.warn(_(b"%s: can't copy - same file\n") % reltarget)
                     return True  # report a failure
                 exists = False
                 samefile = True
 
         if not after and exists or after and already_commited:
             if not opts[b'force']:
                 if already_commited:
                     msg = _(b'%s: not overwriting - file already committed\n')
                     # Check if if the target was added in the parent and the
                     # source already existed in the grandparent.
                     looks_like_copy_in_pctx = abstarget in pctx and any(
                         abssrc in gpctx and abstarget not in gpctx
                         for gpctx in pctx.parents()
                     )
                     if looks_like_copy_in_pctx:
                         if rename:
                             hint = _(
                                 b"('hg rename --at-rev .' to record the rename "
                                 b"in the parent of the working copy)\n"
                             )
                         else:
                             hint = _(
                                 b"('hg copy --at-rev .' to record the copy in "
                                 b"the parent of the working copy)\n"
                             )
                     else:
                         if after:
                             flags = b'--after --force'
                         else:
                             flags = b'--force'
                         if rename:
                             hint = (
                                 _(
                                     b"('hg rename %s' to replace the file by "
                                     b'recording a rename)\n'
                                 )
                                 % flags
                             )
                         else:
                             hint = (
                                 _(
                                     b"('hg copy %s' to replace the file by "
                                     b'recording a copy)\n'
                                 )
                                 % flags
                             )
                 else:
                     msg = _(b'%s: not overwriting - file exists\n')
                     if rename:
                         hint = _(
                             b"('hg rename --after' to record the rename)\n"
                         )
                     else:
                         hint = _(b"('hg copy --after' to record the copy)\n")
                 ui.warn(msg % reltarget)
                 ui.warn(hint)
                 return True  # report a failure
 
         if after:
             if not exists:
                 if rename:
                     ui.warn(
                         _(b'%s: not recording move - %s does not exist\n')
                         % (relsrc, reltarget)
                     )
                 else:
                     ui.warn(
                         _(b'%s: not recording copy - %s does not exist\n')
                         % (relsrc, reltarget)
                     )
                 return True  # report a failure
         elif not dryrun:
             try:
                 if exists:
                     os.unlink(target)
                 targetdir = os.path.dirname(target) or b'.'
                 if not os.path.isdir(targetdir):
                     os.makedirs(targetdir)
                 if samefile:
                     tmp = target + b"~hgrename"
                     os.rename(src, tmp)
                     os.rename(tmp, target)
                 else:
                     # Preserve stat info on renames, not on copies; this matches
                     # Linux CLI behavior.
                     util.copyfile(src, target, copystat=rename)
                 srcexists = True
             except IOError as inst:
                 if inst.errno == errno.ENOENT:
                     ui.warn(_(b'%s: deleted in working directory\n') % relsrc)
                     srcexists = False
                 else:
                     ui.warn(
                         _(b'%s: cannot copy - %s\n')
                         % (relsrc, encoding.strtolocal(inst.strerror))
                     )
                     return True  # report a failure
 
         if ui.verbose or not exact:
             if rename:
                 ui.status(_(b'moving %s to %s\n') % (relsrc, reltarget))
             else:
                 ui.status(_(b'copying %s to %s\n') % (relsrc, reltarget))
 
         targets[abstarget] = abssrc
 
         # fix up dirstate
         scmutil.dirstatecopy(
             ui, repo, ctx, abssrc, abstarget, dryrun=dryrun, cwd=cwd
         )
         if rename and not dryrun:
             if not after and srcexists and not samefile:
                 rmdir = repo.ui.configbool(b'experimental', b'removeemptydirs')
                 repo.wvfs.unlinkpath(abssrc, rmdir=rmdir)
             ctx.forget([abssrc])
 
     # pat: ossep
     # dest ossep
     # srcs: list of (hgsep, hgsep, ossep, bool)
     # return: function that takes hgsep and returns ossep
     def targetpathfn(pat, dest, srcs):
         if os.path.isdir(pat):
             abspfx = pathutil.canonpath(repo.root, cwd, pat)
             abspfx = util.localpath(abspfx)
             if destdirexists:
                 striplen = len(os.path.split(abspfx)[0])
             else:
                 striplen = len(abspfx)
             if striplen:
                 striplen += len(pycompat.ossep)
             res = lambda p: os.path.join(dest, util.localpath(p)[striplen:])
         elif destdirexists:
             res = lambda p: os.path.join(
                 dest, os.path.basename(util.localpath(p))
             )
         else:
             res = lambda p: dest
         return res
 
     # pat: ossep
     # dest ossep
     # srcs: list of (hgsep, hgsep, ossep, bool)
     # return: function that takes hgsep and returns ossep
     def targetpathafterfn(pat, dest, srcs):
         if matchmod.patkind(pat):
             # a mercurial pattern
             res = lambda p: os.path.join(
                 dest, os.path.basename(util.localpath(p))
             )
         else:
             abspfx = pathutil.canonpath(repo.root, cwd, pat)
             if len(abspfx) < len(srcs[0][0]):
                 # A directory. Either the target path contains the last
                 # component of the source path or it does not.
                 def evalpath(striplen):
                     score = 0
                     for s in srcs:
                         t = os.path.join(dest, util.localpath(s[0])[striplen:])
                         if os.path.lexists(t):
                             score += 1
                     return score
 
                 abspfx = util.localpath(abspfx)
                 striplen = len(abspfx)
                 if striplen:
                     striplen += len(pycompat.ossep)
                 if os.path.isdir(os.path.join(dest, os.path.split(abspfx)[1])):
                     score = evalpath(striplen)
                     striplen1 = len(os.path.split(abspfx)[0])
                     if striplen1:
                         striplen1 += len(pycompat.ossep)
                     if evalpath(striplen1) > score:
                         striplen = striplen1
                 res = lambda p: os.path.join(dest, util.localpath(p)[striplen:])
             else:
                 # a file
                 if destdirexists:
                     res = lambda p: os.path.join(
                         dest, os.path.basename(util.localpath(p))
                     )
                 else:
                     res = lambda p: dest
         return res
 
     destdirexists = os.path.isdir(dest) and not os.path.islink(dest)
     if not destdirexists:
         if len(pats) > 1 or matchmod.patkind(pats[0]):
             raise error.InputError(
                 _(
                     b'with multiple sources, destination must be an '
                     b'existing directory'
                 )
             )
         if util.endswithsep(dest):
             raise error.InputError(
                 _(b'destination %s is not a directory') % dest
             )
 
     tfn = targetpathfn
     if after:
         tfn = targetpathafterfn
     copylist = []
     for pat in pats:
         srcs = walkpat(pat)
         if not srcs:
             continue
         copylist.append((tfn(pat, dest, srcs), srcs))
     if not copylist:
         hint = None
         if rename:
             hint = _(b'maybe you meant to use --after --at-rev=.')
         raise error.InputError(_(b'no files to copy'), hint=hint)
 
     errors = 0
     for targetpath, srcs in copylist:
         for abssrc, relsrc, exact in srcs:
             if copyfile(abssrc, relsrc, targetpath(abssrc), exact):
                 errors += 1
 
     return errors != 0
 
 
 ## facility to let extension process additional data into an import patch
 # list of identifier to be executed in order
 extrapreimport = []  # run before commit
 extrapostimport = []  # run after commit
 # mapping from identifier to actual import function
 #
 # 'preimport' are run before the commit is made and are provided the following
 # arguments:
 # - repo: the localrepository instance,
 # - patchdata: data extracted from patch header (cf m.patch.patchheadermap),
 # - extra: the future extra dictionary of the changeset, please mutate it,
 # - opts: the import options.
 # XXX ideally, we would just pass an ctx ready to be computed, that would allow
 # mutation of in memory commit and more. Feel free to rework the code to get
 # there.
 extrapreimportmap = {}
 # 'postimport' are run after the commit is made and are provided the following
 # argument:
 # - ctx: the changectx created by import.
 extrapostimportmap = {}
 
 
 def tryimportone(ui, repo, patchdata, parents, opts, msgs, updatefunc):
     """Utility function used by commands.import to import a single patch
 
     This function is explicitly defined here to help the evolve extension to
     wrap this part of the import logic.
 
     The API is currently a bit ugly because it a simple code translation from
     the import command. Feel free to make it better.
 
     :patchdata: a dictionary containing parsed patch data (such as from
                 ``patch.extract()``)
     :parents: nodes that will be parent of the created commit
     :opts: the full dict of option passed to the import command
     :msgs: list to save commit message to.
            (used in case we need to save it when failing)
     :updatefunc: a function that update a repo to a given node
                  updatefunc(<repo>, <node>)
     """
     # avoid cycle context -> subrepo -> cmdutil
     from . import context
 
     tmpname = patchdata.get(b'filename')
     message = patchdata.get(b'message')
     user = opts.get(b'user') or patchdata.get(b'user')
     date = opts.get(b'date') or patchdata.get(b'date')
     branch = patchdata.get(b'branch')
     nodeid = patchdata.get(b'nodeid')
     p1 = patchdata.get(b'p1')
     p2 = patchdata.get(b'p2')
 
     nocommit = opts.get(b'no_commit')
     importbranch = opts.get(b'import_branch')
     update = not opts.get(b'bypass')
     strip = opts[b"strip"]
     prefix = opts[b"prefix"]
     sim = float(opts.get(b'similarity') or 0)
 
     if not tmpname:
         return None, None, False
 
     rejects = False
 
     cmdline_message = logmessage(ui, opts)
     if cmdline_message:
         # pickup the cmdline msg
         message = cmdline_message
     elif message:
         # pickup the patch msg
         message = message.strip()
     else:
         # launch the editor
         message = None
     ui.debug(b'message:\n%s\n' % (message or b''))
 
     if len(parents) == 1:
         parents.append(repo[nullrev])
     if opts.get(b'exact'):
         if not nodeid or not p1:
             raise error.InputError(_(b'not a Mercurial patch'))
         p1 = repo[p1]
         p2 = repo[p2 or nullrev]
     elif p2:
         try:
             p1 = repo[p1]
             p2 = repo[p2]
             # Without any options, consider p2 only if the
             # patch is being applied on top of the recorded
             # first parent.
             if p1 != parents[0]:
                 p1 = parents[0]
                 p2 = repo[nullrev]
         except error.RepoError:
             p1, p2 = parents
         if p2.rev() == nullrev:
             ui.warn(
                 _(
                     b"warning: import the patch as a normal revision\n"
                     b"(use --exact to import the patch as a merge)\n"
                 )
             )
     else:
         p1, p2 = parents
 
     n = None
     if update:
         if p1 != parents[0]:
             updatefunc(repo, p1.node())
         if p2 != parents[1]:
             repo.setparents(p1.node(), p2.node())
 
         if opts.get(b'exact') or importbranch:
             repo.dirstate.setbranch(branch or b'default')
 
         partial = opts.get(b'partial', False)
         files = set()
         try:
             patch.patch(
                 ui,
                 repo,
                 tmpname,
                 strip=strip,
                 prefix=prefix,
                 files=files,
                 eolmode=None,
                 similarity=sim / 100.0,
             )
         except error.PatchParseError as e:
             raise error.InputError(
                 pycompat.bytestr(e),
                 hint=_(
                     b'check that whitespace in the patch has not been mangled'
                 ),
             )
         except error.PatchApplicationError as e:
             if not partial:
                 raise error.StateError(pycompat.bytestr(e))
             if partial:
                 rejects = True
 
         files = list(files)
         if nocommit:
             if message:
                 msgs.append(message)
         else:
             if opts.get(b'exact') or p2:
                 # If you got here, you either use --force and know what
                 # you are doing or used --exact or a merge patch while
                 # being updated to its first parent.
                 m = None
             else:
                 m = scmutil.matchfiles(repo, files or [])
             editform = mergeeditform(repo[None], b'import.normal')
             if opts.get(b'exact'):
                 editor = None
             else:
                 editor = getcommiteditor(
                     editform=editform, **pycompat.strkwargs(opts)
                 )
             extra = {}
             for idfunc in extrapreimport:
                 extrapreimportmap[idfunc](repo, patchdata, extra, opts)
             overrides = {}
             if partial:
                 overrides[(b'ui', b'allowemptycommit')] = True
             if opts.get(b'secret'):
                 overrides[(b'phases', b'new-commit')] = b'secret'
             with repo.ui.configoverride(overrides, b'import'):
                 n = repo.commit(
                     message, user, date, match=m, editor=editor, extra=extra
                 )
                 for idfunc in extrapostimport:
                     extrapostimportmap[idfunc](repo[n])
     else:
         if opts.get(b'exact') or importbranch:
             branch = branch or b'default'
         else:
             branch = p1.branch()
         store = patch.filestore()
         try:
             files = set()
             try:
                 patch.patchrepo(
                     ui,
                     repo,
                     p1,
                     store,
                     tmpname,
                     strip,
                     prefix,
                     files,
                     eolmode=None,
                 )
             except error.PatchParseError as e:
                 raise error.InputError(
                     stringutil.forcebytestr(e),
                     hint=_(
                         b'check that whitespace in the patch has not been mangled'
                     ),
                 )
             except error.PatchApplicationError as e:
                 raise error.StateError(stringutil.forcebytestr(e))
             if opts.get(b'exact'):
                 editor = None
             else:
                 editor = getcommiteditor(editform=b'import.bypass')
             memctx = context.memctx(
                 repo,
                 (p1.node(), p2.node()),
                 message,
                 files=files,
                 filectxfn=store,
                 user=user,
                 date=date,
                 branch=branch,
                 editor=editor,
             )
 
             overrides = {}
             if opts.get(b'secret'):
                 overrides[(b'phases', b'new-commit')] = b'secret'
             with repo.ui.configoverride(overrides, b'import'):
                 n = memctx.commit()
         finally:
             store.close()
     if opts.get(b'exact') and nocommit:
         # --exact with --no-commit is still useful in that it does merge
         # and branch bits
         ui.warn(_(b"warning: can't check exact import with --no-commit\n"))
     elif opts.get(b'exact') and (not n or hex(n) != nodeid):
         raise error.Abort(_(b'patch is damaged or loses information'))
     msg = _(b'applied to working directory')
     if n:
         # i18n: refers to a short changeset id
         msg = _(b'created %s') % short(n)
     return msg, n, rejects
 
 
 # facility to let extensions include additional data in an exported patch
 # list of identifiers to be executed in order
 extraexport = []
 # mapping from identifier to actual export function
 # function as to return a string to be added to the header or None
 # it is given two arguments (sequencenumber, changectx)
 extraexportmap = {}
 
 
 def _exportsingle(repo, ctx, fm, match, switch_parent, seqno, diffopts):
     node = scmutil.binnode(ctx)
     parents = [p.node() for p in ctx.parents() if p]
     branch = ctx.branch()
     if switch_parent:
         parents.reverse()
 
     if parents:
         prev = parents[0]
     else:
         prev = repo.nullid
 
     fm.context(ctx=ctx)
     fm.plain(b'# HG changeset patch\n')
     fm.write(b'user', b'# User %s\n', ctx.user())
     fm.plain(b'# Date %d %d\n' % ctx.date())
     fm.write(b'date', b'#      %s\n', fm.formatdate(ctx.date()))
     fm.condwrite(
         branch and branch != b'default', b'branch', b'# Branch %s\n', branch
     )
     fm.write(b'node', b'# Node ID %s\n', hex(node))
     fm.plain(b'# Parent  %s\n' % hex(prev))
     if len(parents) > 1:
         fm.plain(b'# Parent  %s\n' % hex(parents[1]))
     fm.data(parents=fm.formatlist(pycompat.maplist(hex, parents), name=b'node'))
 
     # TODO: redesign extraexportmap function to support formatter
     for headerid in extraexport:
         header = extraexportmap[headerid](seqno, ctx)
         if header is not None:
             fm.plain(b'# %s\n' % header)
 
     fm.write(b'desc', b'%s\n', ctx.description().rstrip())
     fm.plain(b'\n')
 
     if fm.isplain():
         chunkiter = patch.diffui(repo, prev, node, match, opts=diffopts)
         for chunk, label in chunkiter:
             fm.plain(chunk, label=label)
     else:
         chunkiter = patch.diff(repo, prev, node, match, opts=diffopts)
         # TODO: make it structured?
         fm.data(diff=b''.join(chunkiter))
 
 
 def _exportfile(repo, revs, fm, dest, switch_parent, diffopts, match):
     """Export changesets to stdout or a single file"""
     for seqno, rev in enumerate(revs, 1):
         ctx = repo[rev]
         if not dest.startswith(b'<'):
             repo.ui.note(b"%s\n" % dest)
         fm.startitem()
         _exportsingle(repo, ctx, fm, match, switch_parent, seqno, diffopts)
 
 
 def _exportfntemplate(
     repo, revs, basefm, fntemplate, switch_parent, diffopts, match
 ):
     """Export changesets to possibly multiple files"""
     total = len(revs)
     revwidth = max(len(str(rev)) for rev in revs)
     filemap = util.sortdict()  # filename: [(seqno, rev), ...]
 
     for seqno, rev in enumerate(revs, 1):
         ctx = repo[rev]
         dest = makefilename(
             ctx, fntemplate, total=total, seqno=seqno, revwidth=revwidth
         )
         filemap.setdefault(dest, []).append((seqno, rev))
 
     for dest in filemap:
         with formatter.maybereopen(basefm, dest) as fm:
             repo.ui.note(b"%s\n" % dest)
             for seqno, rev in filemap[dest]:
                 fm.startitem()
                 ctx = repo[rev]
                 _exportsingle(
                     repo, ctx, fm, match, switch_parent, seqno, diffopts
                 )
 
 
 def _prefetchchangedfiles(repo, revs, match):
     allfiles = set()
     for rev in revs:
         for file in repo[rev].files():
             if not match or match(file):
                 allfiles.add(file)
     match = scmutil.matchfiles(repo, allfiles)
     revmatches = [(rev, match) for rev in revs]
     scmutil.prefetchfiles(repo, revmatches)
 
 
 def export(
     repo,
     revs,
     basefm,
     fntemplate=b'hg-%h.patch',
     switch_parent=False,
     opts=None,
     match=None,
 ):
     """export changesets as hg patches
 
     Args:
       repo: The repository from which we're exporting revisions.
       revs: A list of revisions to export as revision numbers.
       basefm: A formatter to which patches should be written.
       fntemplate: An optional string to use for generating patch file names.
       switch_parent: If True, show diffs against second parent when not nullid.
                      Default is false, which always shows diff against p1.
       opts: diff options to use for generating the patch.
       match: If specified, only export changes to files matching this matcher.
 
     Returns:
       Nothing.
 
     Side Effect:
       "HG Changeset Patch" data is emitted to one of the following
       destinations:
         fntemplate specified: Each rev is written to a unique file named using
                             the given template.
         Otherwise: All revs will be written to basefm.
     """
     _prefetchchangedfiles(repo, revs, match)
 
     if not fntemplate:
         _exportfile(
             repo, revs, basefm, b'<unnamed>', switch_parent, opts, match
         )
     else:
         _exportfntemplate(
             repo, revs, basefm, fntemplate, switch_parent, opts, match
         )
 
 
 def exportfile(repo, revs, fp, switch_parent=False, opts=None, match=None):
     """Export changesets to the given file stream"""
     _prefetchchangedfiles(repo, revs, match)
 
     dest = getattr(fp, 'name', b'<unnamed>')
     with formatter.formatter(repo.ui, fp, b'export', {}) as fm:
         _exportfile(repo, revs, fm, dest, switch_parent, opts, match)
 
 
 def showmarker(fm, marker, index=None):
     """utility function to display obsolescence marker in a readable way
 
     To be used by debug function."""
     if index is not None:
         fm.write(b'index', b'%i ', index)
     fm.write(b'prednode', b'%s ', hex(marker.prednode()))
     succs = marker.succnodes()
     fm.condwrite(
         succs,
         b'succnodes',
         b'%s ',
         fm.formatlist(map(hex, succs), name=b'node'),
     )
     fm.write(b'flag', b'%X ', marker.flags())
     parents = marker.parentnodes()
     if parents is not None:
         fm.write(
             b'parentnodes',
             b'{%s} ',
             fm.formatlist(map(hex, parents), name=b'node', sep=b', '),
         )
     fm.write(b'date', b'(%s) ', fm.formatdate(marker.date()))
     meta = marker.metadata().copy()
     meta.pop(b'date', None)
     smeta = pycompat.rapply(pycompat.maybebytestr, meta)
     fm.write(
         b'metadata', b'{%s}', fm.formatdict(smeta, fmt=b'%r: %r', sep=b', ')
     )
     fm.plain(b'\n')
 
 
 def finddate(ui, repo, date):
     """Find the tipmost changeset that matches the given date spec"""
     mrevs = repo.revs(b'date(%s)', date)
     try:
         rev = mrevs.max()
     except ValueError:
         raise error.InputError(_(b"revision matching date not found"))
 
     ui.status(
         _(b"found revision %d from %s\n")
         % (rev, dateutil.datestr(repo[rev].date()))
     )
     return b'%d' % rev
 
 
 def add(ui, repo, match, prefix, uipathfn, explicitonly, **opts):
     bad = []
 
     badfn = lambda x, y: bad.append(x) or match.bad(x, y)
     names = []
     wctx = repo[None]
     cca = None
     abort, warn = scmutil.checkportabilityalert(ui)
     if abort or warn:
         cca = scmutil.casecollisionauditor(ui, abort, repo.dirstate)
 
     match = repo.narrowmatch(match, includeexact=True)
     badmatch = matchmod.badmatch(match, badfn)
     dirstate = repo.dirstate
     # We don't want to just call wctx.walk here, since it would return a lot of
     # clean files, which we aren't interested in and takes time.
     for f in sorted(
         dirstate.walk(
             badmatch,
             subrepos=sorted(wctx.substate),
             unknown=True,
             ignored=False,
             full=False,
         )
     ):
         exact = match.exact(f)
         if exact or not explicitonly and f not in wctx and repo.wvfs.lexists(f):
             if cca:
                 cca(f)
             names.append(f)
             if ui.verbose or not exact:
                 ui.status(
                     _(b'adding %s\n') % uipathfn(f), label=b'ui.addremove.added'
                 )
 
     for subpath in sorted(wctx.substate):
         sub = wctx.sub(subpath)
         try:
             submatch = matchmod.subdirmatcher(subpath, match)
             subprefix = repo.wvfs.reljoin(prefix, subpath)
             subuipathfn = scmutil.subdiruipathfn(subpath, uipathfn)
             if opts.get('subrepos'):
                 bad.extend(
                     sub.add(ui, submatch, subprefix, subuipathfn, False, **opts)
                 )
             else:
                 bad.extend(
                     sub.add(ui, submatch, subprefix, subuipathfn, True, **opts)
                 )
         except error.LookupError:
             ui.status(
                 _(b"skipping missing subrepository: %s\n") % uipathfn(subpath)
             )
 
     if not opts.get('dry_run'):
         rejected = wctx.add(names, prefix)
         bad.extend(f for f in rejected if f in match.files())
     return bad
 
 
 def addwebdirpath(repo, serverpath, webconf):
     webconf[serverpath] = repo.root
     repo.ui.debug(b'adding %s = %s\n' % (serverpath, repo.root))
 
     for r in repo.revs(b'filelog("path:.hgsub")'):
         ctx = repo[r]
         for subpath in ctx.substate:
             ctx.sub(subpath).addwebdirpath(serverpath, webconf)
 
 
 def forget(
     ui, repo, match, prefix, uipathfn, explicitonly, dryrun, interactive
 ):
     if dryrun and interactive:
         raise error.InputError(
             _(b"cannot specify both --dry-run and --interactive")
         )
     bad = []
     badfn = lambda x, y: bad.append(x) or match.bad(x, y)
     wctx = repo[None]
     forgot = []
 
     s = repo.status(match=matchmod.badmatch(match, badfn), clean=True)
     forget = sorted(s.modified + s.added + s.deleted + s.clean)
     if explicitonly:
         forget = [f for f in forget if match.exact(f)]
 
     for subpath in sorted(wctx.substate):
         sub = wctx.sub(subpath)
         submatch = matchmod.subdirmatcher(subpath, match)
         subprefix = repo.wvfs.reljoin(prefix, subpath)
         subuipathfn = scmutil.subdiruipathfn(subpath, uipathfn)
         try:
             subbad, subforgot = sub.forget(
                 submatch,
                 subprefix,
                 subuipathfn,
                 dryrun=dryrun,
                 interactive=interactive,
             )
             bad.extend([subpath + b'/' + f for f in subbad])
             forgot.extend([subpath + b'/' + f for f in subforgot])
         except error.LookupError:
             ui.status(
                 _(b"skipping missing subrepository: %s\n") % uipathfn(subpath)
             )
 
     if not explicitonly:
         for f in match.files():
             if f not in repo.dirstate and not repo.wvfs.isdir(f):
                 if f not in forgot:
                     if repo.wvfs.exists(f):
                         # Don't complain if the exact case match wasn't given.
                         # But don't do this until after checking 'forgot', so
                         # that subrepo files aren't normalized, and this op is
                         # purely from data cached by the status walk above.
                         if repo.dirstate.normalize(f) in repo.dirstate:
                             continue
                         ui.warn(
                             _(
                                 b'not removing %s: '
                                 b'file is already untracked\n'
                             )
                             % uipathfn(f)
                         )
                     bad.append(f)
 
     if interactive:
         responses = _(
             b'[Ynsa?]'
             b'$$ &Yes, forget this file'
             b'$$ &No, skip this file'
             b'$$ &Skip remaining files'
             b'$$ Include &all remaining files'
             b'$$ &? (display help)'
         )
         for filename in forget[:]:
             r = ui.promptchoice(
                 _(b'forget %s %s') % (uipathfn(filename), responses)
             )
             if r == 4:  # ?
                 while r == 4:
                     for c, t in ui.extractchoices(responses)[1]:
                         ui.write(b'%s - %s\n' % (c, encoding.lower(t)))
                     r = ui.promptchoice(
                         _(b'forget %s %s') % (uipathfn(filename), responses)
                     )
             if r == 0:  # yes
                 continue
             elif r == 1:  # no
                 forget.remove(filename)
             elif r == 2:  # Skip
                 fnindex = forget.index(filename)
                 del forget[fnindex:]
                 break
             elif r == 3:  # All
                 break
 
     for f in forget:
         if ui.verbose or not match.exact(f) or interactive:
             ui.status(
                 _(b'removing %s\n') % uipathfn(f), label=b'ui.addremove.removed'
             )
 
     if not dryrun:
         rejected = wctx.forget(forget, prefix)
         bad.extend(f for f in rejected if f in match.files())
         forgot.extend(f for f in forget if f not in rejected)
     return bad, forgot
 
 
 def files(ui, ctx, m, uipathfn, fm, fmt, subrepos):
     ret = 1
 
     needsfctx = ui.verbose or {b'size', b'flags'} & fm.datahint()
     if fm.isplain() and not needsfctx:
         # Fast path. The speed-up comes from skipping the formatter, and batching
         # calls to ui.write.
         buf = []
         for f in ctx.matches(m):
             buf.append(fmt % uipathfn(f))
             if len(buf) > 100:
                 ui.write(b''.join(buf))
                 del buf[:]
             ret = 0
         if buf:
             ui.write(b''.join(buf))
     else:
         for f in ctx.matches(m):
             fm.startitem()
             fm.context(ctx=ctx)
             if needsfctx:
                 fc = ctx[f]
                 fm.write(b'size flags', b'% 10d % 1s ', fc.size(), fc.flags())
             fm.data(path=f)
             fm.plain(fmt % uipathfn(f))
             ret = 0
 
     for subpath in sorted(ctx.substate):
         submatch = matchmod.subdirmatcher(subpath, m)
         subuipathfn = scmutil.subdiruipathfn(subpath, uipathfn)
         if subrepos or m.exact(subpath) or any(submatch.files()):
             sub = ctx.sub(subpath)
             try:
                 recurse = m.exact(subpath) or subrepos
                 if (
                     sub.printfiles(ui, submatch, subuipathfn, fm, fmt, recurse)
                     == 0
                 ):
                     ret = 0
             except error.LookupError:
                 ui.status(
                     _(b"skipping missing subrepository: %s\n")
                     % uipathfn(subpath)
                 )
 
     return ret
 
 
 def remove(
     ui, repo, m, prefix, uipathfn, after, force, subrepos, dryrun, warnings=None
 ):
     ret = 0
     s = repo.status(match=m, clean=True)
     modified, added, deleted, clean = s.modified, s.added, s.deleted, s.clean
 
     wctx = repo[None]
 
     if warnings is None:
         warnings = []
         warn = True
     else:
         warn = False
 
     subs = sorted(wctx.substate)
     progress = ui.makeprogress(
         _(b'searching'), total=len(subs), unit=_(b'subrepos')
     )
     for subpath in subs:
         submatch = matchmod.subdirmatcher(subpath, m)
         subprefix = repo.wvfs.reljoin(prefix, subpath)
         subuipathfn = scmutil.subdiruipathfn(subpath, uipathfn)
         if subrepos or m.exact(subpath) or any(submatch.files()):
             progress.increment()
             sub = wctx.sub(subpath)
             try:
                 if sub.removefiles(
                     submatch,
                     subprefix,
                     subuipathfn,
                     after,
                     force,
                     subrepos,
                     dryrun,
                     warnings,
                 ):
                     ret = 1
             except error.LookupError:
                 warnings.append(
                     _(b"skipping missing subrepository: %s\n")
                     % uipathfn(subpath)
                 )
     progress.complete()
 
     # warn about failure to delete explicit files/dirs
     deleteddirs = pathutil.dirs(deleted)
     files = m.files()
     progress = ui.makeprogress(
         _(b'deleting'), total=len(files), unit=_(b'files')
     )
     for f in files:
 
         def insubrepo():
             for subpath in wctx.substate:
                 if f.startswith(subpath + b'/'):
                     return True
             return False
 
         progress.increment()
         isdir = f in deleteddirs or wctx.hasdir(f)
         if f in repo.dirstate or isdir or f == b'.' or insubrepo() or f in subs:
             continue
 
         if repo.wvfs.exists(f):
             if repo.wvfs.isdir(f):
                 warnings.append(
                     _(b'not removing %s: no tracked files\n') % uipathfn(f)
                 )
             else:
                 warnings.append(
                     _(b'not removing %s: file is untracked\n') % uipathfn(f)
                 )
         # missing files will generate a warning elsewhere
         ret = 1
     progress.complete()
 
     if force:
         list = modified + deleted + clean + added
     elif after:
         list = deleted
         remaining = modified + added + clean
         progress = ui.makeprogress(
             _(b'skipping'), total=len(remaining), unit=_(b'files')
         )
         for f in remaining:
             progress.increment()
             if ui.verbose or (f in files):
                 warnings.append(
                     _(b'not removing %s: file still exists\n') % uipathfn(f)
                 )
             ret = 1
         progress.complete()
     else:
         list = deleted + clean
         progress = ui.makeprogress(
             _(b'skipping'), total=(len(modified) + len(added)), unit=_(b'files')
         )
         for f in modified:
             progress.increment()
             warnings.append(
                 _(
                     b'not removing %s: file is modified (use -f'
                     b' to force removal)\n'
                 )
                 % uipathfn(f)
             )
             ret = 1
         for f in added:
             progress.increment()
             warnings.append(
                 _(
                     b"not removing %s: file has been marked for add"
                     b" (use 'hg forget' to undo add)\n"
                 )
                 % uipathfn(f)
             )
             ret = 1
         progress.complete()
 
     list = sorted(list)
     progress = ui.makeprogress(
         _(b'deleting'), total=len(list), unit=_(b'files')
     )
     for f in list:
         if ui.verbose or not m.exact(f):
             progress.increment()
             ui.status(
                 _(b'removing %s\n') % uipathfn(f), label=b'ui.addremove.removed'
             )
     progress.complete()
 
     if not dryrun:
         with repo.wlock():
             if not after:
                 for f in list:
                     if f in added:
                         continue  # we never unlink added files on remove
                     rmdir = repo.ui.configbool(
                         b'experimental', b'removeemptydirs'
                     )
                     repo.wvfs.unlinkpath(f, ignoremissing=True, rmdir=rmdir)
             repo[None].forget(list)
 
     if warn:
         for warning in warnings:
             ui.warn(warning)
 
     return ret
 
 
 def _catfmtneedsdata(fm):
     return not fm.datahint() or b'data' in fm.datahint()
 
 
 def _updatecatformatter(fm, ctx, matcher, path, decode):
     """Hook for adding data to the formatter used by ``hg cat``.
 
     Extensions (e.g., lfs) can wrap this to inject keywords/data, but must call
     this method first."""
 
     # data() can be expensive to fetch (e.g. lfs), so don't fetch it if it
     # wasn't requested.
     data = b''
     if _catfmtneedsdata(fm):
         data = ctx[path].data()
         if decode:
             data = ctx.repo().wwritedata(path, data)
     fm.startitem()
     fm.context(ctx=ctx)
     fm.write(b'data', b'%s', data)
     fm.data(path=path)
 
 
 def cat(ui, repo, ctx, matcher, basefm, fntemplate, prefix, **opts):
     err = 1
     opts = pycompat.byteskwargs(opts)
 
     def write(path):
         filename = None
         if fntemplate:
             filename = makefilename(
                 ctx, fntemplate, pathname=os.path.join(prefix, path)
             )
             # attempt to create the directory if it does not already exist
             try:
                 os.makedirs(os.path.dirname(filename))
             except OSError:
                 pass
         with formatter.maybereopen(basefm, filename) as fm:
             _updatecatformatter(fm, ctx, matcher, path, opts.get(b'decode'))
 
     # Automation often uses hg cat on single files, so special case it
     # for performance to avoid the cost of parsing the manifest.
     if len(matcher.files()) == 1 and not matcher.anypats():
         file = matcher.files()[0]
         mfl = repo.manifestlog
         mfnode = ctx.manifestnode()
         try:
             if mfnode and mfl[mfnode].find(file)[0]:
                 if _catfmtneedsdata(basefm):
                     scmutil.prefetchfiles(repo, [(ctx.rev(), matcher)])
                 write(file)
                 return 0
         except KeyError:
             pass
 
     if _catfmtneedsdata(basefm):
         scmutil.prefetchfiles(repo, [(ctx.rev(), matcher)])
 
     for abs in ctx.walk(matcher):
         write(abs)
         err = 0
 
     uipathfn = scmutil.getuipathfn(repo, legacyrelativevalue=True)
     for subpath in sorted(ctx.substate):
         sub = ctx.sub(subpath)
         try:
             submatch = matchmod.subdirmatcher(subpath, matcher)
             subprefix = os.path.join(prefix, subpath)
             if not sub.cat(
                 submatch,
                 basefm,
                 fntemplate,
                 subprefix,
                 **pycompat.strkwargs(opts)
             ):
                 err = 0
         except error.RepoLookupError:
             ui.status(
                 _(b"skipping missing subrepository: %s\n") % uipathfn(subpath)
             )
 
     return err
 
 
+class _AddRemoveContext:
+    """a small (hacky) context to deal with lazy opening of context
+
+    This is to be used in the `commit` function right below. This deals with
+    lazily open a `changing_files` context inside a `transaction` that span the
+    full commit operation.
+
+    We need :
+    - a `changing_files` context to wrap the dirstate change within the
+      "addremove" operation,
+    - a transaction to make sure these change are not written right after the
+      addremove, but when the commit operation succeed.
+
+    However it get complicated because:
+    - opening a transaction "this early" shuffle hooks order, especially the
+      `precommit` one happening after the `pretxtopen` one which I am not too
+      enthusiastic about.
+    - the `mq` extensions + the `record` extension stacks many layers of call
+      to implement `qrefresh --interactive` and this result with `mq` calling a
+      `strip` in the middle of this function. Which prevent the existence of
+      transaction wrapping all of its function code. (however, `qrefresh` never
+      call the `addremove` bits.
+    - the largefile extensions (and maybe other extensions?) wraps `addremove`
+      so slicing `addremove` in smaller bits is a complex endeavour.
+
+    So I eventually took a this shortcut that open the transaction if we
+    actually needs it, not disturbing much of the rest of the code.
+
+    It will result in some hooks order change for `hg commit --addremove`,
+    however it seems a corner case enough to ignore that for now (hopefully).
+
+    Notes that None of the above problems seems insurmountable, however I have
+    been fighting with this specific piece of code for a couple of day already
+    and I need a solution to keep moving forward on the bigger work around
+    `changing_files` context that is being introduced at the same time as this
+    hack.
+
+    Each problem seems to have a solution:
+    - the hook order issue could be solved by refactoring the many-layer stack
+      that currently composes a commit and calling them earlier,
+    - the mq issue could be solved by refactoring `mq` so that the final strip
+      is done after transaction closure. Be warned that the mq code is quite
+      antic however.
+    - large-file could be reworked in parallel of the `addremove` to be
+      friendlier to this.
+
+    However each of these tasks are too much a diversion right now. In addition
+    they will be much easier to undertake when the `changing_files` dust has
+    settled."""
+
+    def __init__(self, repo):
+        self._repo = repo
+        self._transaction = None
+        self._dirstate_context = None
+        self._state = None
+
+    def __enter__(self):
+        assert self._state is None
+        self._state = True
+        return self
+
+    def open_transaction(self):
+        """open a `transaction` and `changing_files` context
+
+        Call this when you know that change to the dirstate will be needed and
+        we need to open the transaction early
+
+        This will also open the dirstate `changing_files` context, so you should
+        call `close_dirstate_context` when the distate changes are done.
+        """
+        assert self._state is not None
+        if self._transaction is None:
+            self._transaction = self._repo.transaction(b'commit')
+            self._transaction.__enter__()
+        if self._dirstate_context is None:
+            self._dirstate_context = self._repo.dirstate.changing_files(
+                self._repo
+            )
+            self._dirstate_context.__enter__()
+
+    def close_dirstate_context(self):
+        """close the change_files if any
+
+        Call this after the (potential) `open_transaction` call to close the
+        (potential) changing_files context.
+        """
+        if self._dirstate_context is not None:
+            self._dirstate_context.__exit__(None, None, None)
+            self._dirstate_context = None
+
+    def __exit__(self, *args):
+        if self._dirstate_context is not None:
+            self._dirstate_context.__exit__(*args)
+        if self._transaction is not None:
+            self._transaction.__exit__(*args)
+
+
 def commit(ui, repo, commitfunc, pats, opts):
     '''commit the specified files or all outstanding changes'''
     date = opts.get(b'date')
     if date:
         opts[b'date'] = dateutil.parsedate(date)
 
-    dsguard = None
-    # extract addremove carefully -- this function can be called from a command
-    # that doesn't support addremove
-    if opts.get(b'addremove'):
-        dsguard = dirstateguard.dirstateguard(repo, b'commit')
-    with dsguard or util.nullcontextmanager():
+    with repo.wlock(), repo.lock():
         message = logmessage(ui, opts)
         matcher = scmutil.match(repo[None], pats, opts)
-        if True:
+
+        with _AddRemoveContext(repo) as c:
             # extract addremove carefully -- this function can be called from a
             # command that doesn't support addremove
             if opts.get(b'addremove'):
                 relative = scmutil.anypats(pats, opts)
                 uipathfn = scmutil.getuipathfn(
                     repo,
                     legacyrelativevalue=relative,
                 )
                 r = scmutil.addremove(
                     repo,
                     matcher,
                     b"",
                     uipathfn,
                     opts,
+                    open_tr=c.open_transaction,
                 )
                 m = _(b"failed to mark all new/missing files as added/removed")
                 if r != 0:
                     raise error.Abort(m)
-
+            c.close_dirstate_context()
             return commitfunc(ui, repo, message, matcher, opts)
 
 
 def samefile(f, ctx1, ctx2):
     if f in ctx1.manifest():
         a = ctx1.filectx(f)
         if f in ctx2.manifest():
             b = ctx2.filectx(f)
             return not a.cmp(b) and a.flags() == b.flags()
         else:
             return False
     else:
         return f not in ctx2.manifest()
 
 
 def amend(ui, repo, old, extra, pats, opts):
     # avoid cycle context -> subrepo -> cmdutil
     from . import context
 
     # amend will reuse the existing user if not specified, but the obsolete
     # marker creation requires that the current user's name is specified.
     if obsolete.isenabled(repo, obsolete.createmarkersopt):
         ui.username()  # raise exception if username not set
 
     ui.note(_(b'amending changeset %s\n') % old)
     base = old.p1()
 
     with repo.wlock(), repo.lock(), repo.transaction(b'amend'):
         # Participating changesets:
         #
         # wctx     o - workingctx that contains changes from working copy
         #          |   to go into amending commit
         #          |
         # old      o - changeset to amend
         #          |
         # base     o - first parent of the changeset to amend
         wctx = repo[None]
 
         # Copy to avoid mutating input
         extra = extra.copy()
         # Update extra dict from amended commit (e.g. to preserve graft
         # source)
         extra.update(old.extra())
 
         # Also update it from the from the wctx
         extra.update(wctx.extra())
 
         # date-only change should be ignored?
         datemaydiffer = resolve_commit_options(ui, opts)
         opts = pycompat.byteskwargs(opts)
 
         date = old.date()
         if opts.get(b'date'):
             date = dateutil.parsedate(opts.get(b'date'))
         user = opts.get(b'user') or old.user()
 
         if len(old.parents()) > 1:
             # ctx.files() isn't reliable for merges, so fall back to the
             # slower repo.status() method
             st = base.status(old)
             files = set(st.modified) | set(st.added) | set(st.removed)
         else:
             files = set(old.files())
 
         # add/remove the files to the working copy if the "addremove" option
         # was specified.
         matcher = scmutil.match(wctx, pats, opts)
         relative = scmutil.anypats(pats, opts)
         uipathfn = scmutil.getuipathfn(repo, legacyrelativevalue=relative)
         if opts.get(b'addremove') and scmutil.addremove(
             repo, matcher, b"", uipathfn, opts
         ):
             raise error.Abort(
                 _(b"failed to mark all new/missing files as added/removed")
             )
 
         # Check subrepos. This depends on in-place wctx._status update in
         # subrepo.precommit(). To minimize the risk of this hack, we do
         # nothing if .hgsub does not exist.
         if b'.hgsub' in wctx or b'.hgsub' in old:
             subs, commitsubs, newsubstate = subrepoutil.precommit(
                 ui, wctx, wctx._status, matcher
             )
             # amend should abort if commitsubrepos is enabled
             assert not commitsubs
             if subs:
                 subrepoutil.writestate(repo, newsubstate)
 
         ms = mergestatemod.mergestate.read(repo)
         mergeutil.checkunresolved(ms)
 
         filestoamend = {f for f in wctx.files() if matcher(f)}
 
         changes = len(filestoamend) > 0
         changeset_copies = (
             repo.ui.config(b'experimental', b'copies.read-from')
             != b'filelog-only'
         )
         # If there are changes to amend or if copy information needs to be read
         # from the changeset extras, we cannot take the fast path of using
         # filectxs from the old commit.
         if changes or changeset_copies:
             # Recompute copies (avoid recording a -> b -> a)
             copied = copies.pathcopies(base, wctx)
             if old.p2():
                 copied.update(copies.pathcopies(old.p2(), wctx))
 
             # Prune files which were reverted by the updates: if old
             # introduced file X and the file was renamed in the working
             # copy, then those two files are the same and
             # we can discard X from our list of files. Likewise if X
             # was removed, it's no longer relevant. If X is missing (aka
             # deleted), old X must be preserved.
             files.update(filestoamend)
             files = [
                 f
                 for f in files
                 if (f not in filestoamend or not samefile(f, wctx, base))
             ]
 
             def filectxfn(repo, ctx_, path):
                 try:
                     # If the file being considered is not amongst the files
                     # to be amended, we should use the file context from the
                     # old changeset. This avoids issues when only some files in
                     # the working copy are being amended but there are also
                     # changes to other files from the old changeset.
                     if path in filestoamend:
                         # Return None for removed files.
                         if path in wctx.removed():
                             return None
                         fctx = wctx[path]
                     else:
                         fctx = old.filectx(path)
                     flags = fctx.flags()
                     mctx = context.memfilectx(
                         repo,
                         ctx_,
                         fctx.path(),
                         fctx.data(),
                         islink=b'l' in flags,
                         isexec=b'x' in flags,
                         copysource=copied.get(path),
                     )
                     return mctx
                 except KeyError:
                     return None
 
         else:
             ui.note(_(b'copying changeset %s to %s\n') % (old, base))
 
             # Use version of files as in the old cset
             def filectxfn(repo, ctx_, path):
                 try:
                     return old.filectx(path)
                 except KeyError:
                     return None
 
         # See if we got a message from -m or -l, if not, open the editor with
         # the message of the changeset to amend.
         message = logmessage(ui, opts)
 
         editform = mergeeditform(old, b'commit.amend')
 
         if not message:
             message = old.description()
             # Default if message isn't provided and --edit is not passed is to
             # invoke editor, but allow --no-edit. If somehow we don't have any
             # description, let's always start the editor.
             doedit = not message or opts.get(b'edit') in [True, None]
         else:
             # Default if message is provided is to not invoke editor, but allow
             # --edit.
             doedit = opts.get(b'edit') is True
         editor = getcommiteditor(edit=doedit, editform=editform)
 
         pureextra = extra.copy()
         extra[b'amend_source'] = old.hex()
 
         new = context.memctx(
             repo,
             parents=[base.node(), old.p2().node()],
             text=message,
             files=files,
             filectxfn=filectxfn,
             user=user,
             date=date,
             extra=extra,
             editor=editor,
         )
 
         newdesc = changelog.stripdesc(new.description())
         if (
             (not changes)
             and newdesc == old.description()
             and user == old.user()
             and (date == old.date() or datemaydiffer)
             and pureextra == old.extra()
         ):
             # nothing changed. continuing here would create a new node
             # anyway because of the amend_source noise.
             #
             # This not what we expect from amend.
             return old.node()
 
         commitphase = None
         if opts.get(b'secret'):
             commitphase = phases.secret
         elif opts.get(b'draft'):
             commitphase = phases.draft
         newid = repo.commitctx(new)
         ms.reset()
 
         with repo.dirstate.changing_parents(repo):
             # Reroute the working copy parent to the new changeset
             repo.setparents(newid, repo.nullid)
 
             # Fixing the dirstate because localrepo.commitctx does not update
             # it. This is rather convenient because we did not need to update
             # the dirstate for all the files in the new commit which commitctx
             # could have done if it updated the dirstate. Now, we can
             # selectively update the dirstate only for the amended files.
             dirstate = repo.dirstate
 
             # Update the state of the files which were added and modified in the
             # amend to "normal" in the dirstate. We need to use "normallookup" since
             # the files may have changed since the command started; using "normal"
             # would mark them as clean but with uncommitted contents.
             normalfiles = set(wctx.modified() + wctx.added()) & filestoamend
             for f in normalfiles:
                 dirstate.update_file(
                     f, p1_tracked=True, wc_tracked=True, possibly_dirty=True
                 )
 
             # Update the state of files which were removed in the amend
             # to "removed" in the dirstate.
             removedfiles = set(wctx.removed()) & filestoamend
             for f in removedfiles:
                 dirstate.update_file(f, p1_tracked=False, wc_tracked=False)
 
         mapping = {old.node(): (newid,)}
         obsmetadata = None
         if opts.get(b'note'):
             obsmetadata = {b'note': encoding.fromlocal(opts[b'note'])}
         backup = ui.configbool(b'rewrite', b'backup-bundle')
         scmutil.cleanupnodes(
             repo,
             mapping,
             b'amend',
             metadata=obsmetadata,
             fixphase=True,
             targetphase=commitphase,
             backup=backup,
         )
 
     return newid
 
 
 def commiteditor(repo, ctx, subs, editform=b''):
     if ctx.description():
         return ctx.description()
     return commitforceeditor(
         repo, ctx, subs, editform=editform, unchangedmessagedetection=True
     )
 
 
 def commitforceeditor(
     repo,
     ctx,
     subs,
     finishdesc=None,
     extramsg=None,
     editform=b'',
     unchangedmessagedetection=False,
 ):
     if not extramsg:
         extramsg = _(b"Leave message empty to abort commit.")
 
     forms = [e for e in editform.split(b'.') if e]
     forms.insert(0, b'changeset')
     templatetext = None
     while forms:
         ref = b'.'.join(forms)
         if repo.ui.config(b'committemplate', ref):
             templatetext = committext = buildcommittemplate(
                 repo, ctx, subs, extramsg, ref
             )
             break
         forms.pop()
     else:
         committext = buildcommittext(repo, ctx, subs, extramsg)
 
     # run editor in the repository root
     olddir = encoding.getcwd()
     os.chdir(repo.root)
 
     # make in-memory changes visible to external process
     tr = repo.currenttransaction()
     repo.dirstate.write(tr)
     pending = tr and tr.writepending() and repo.root
 
     editortext = repo.ui.edit(
         committext,
         ctx.user(),
         ctx.extra(),
         editform=editform,
         pending=pending,
         repopath=repo.path,
         action=b'commit',
     )
     text = editortext
 
     # strip away anything below this special string (used for editors that want
     # to display the diff)
     stripbelow = re.search(_linebelow, text, flags=re.MULTILINE)
     if stripbelow:
         text = text[: stripbelow.start()]
 
     text = re.sub(b"(?m)^HG:.*(\n|$)", b"", text)
     os.chdir(olddir)
 
     if finishdesc:
         text = finishdesc(text)
     if not text.strip():
         raise error.InputError(_(b"empty commit message"))
     if unchangedmessagedetection and editortext == templatetext:
         raise error.InputError(_(b"commit message unchanged"))
 
     return text
 
 
 def buildcommittemplate(repo, ctx, subs, extramsg, ref):
     ui = repo.ui
     spec = formatter.reference_templatespec(ref)
     t = logcmdutil.changesettemplater(ui, repo, spec)
     t.t.cache.update(
         (k, templater.unquotestring(v))
         for k, v in repo.ui.configitems(b'committemplate')
     )
 
     if not extramsg:
         extramsg = b''  # ensure that extramsg is string
 
     ui.pushbuffer()
     t.show(ctx, extramsg=extramsg)
     return ui.popbuffer()
 
 
 def hgprefix(msg):
     return b"\n".join([b"HG: %s" % a for a in msg.split(b"\n") if a])
 
 
 def buildcommittext(repo, ctx, subs, extramsg):
     edittext = []
     modified, added, removed = ctx.modified(), ctx.added(), ctx.removed()
     if ctx.description():
         edittext.append(ctx.description())
     edittext.append(b"")
     edittext.append(b"")  # Empty line between message and comments.
     edittext.append(
         hgprefix(
             _(
                 b"Enter commit message."
                 b"  Lines beginning with 'HG:' are removed."
             )
         )
     )
     edittext.append(hgprefix(extramsg))
     edittext.append(b"HG: --")
     edittext.append(hgprefix(_(b"user: %s") % ctx.user()))
     if ctx.p2():
         edittext.append(hgprefix(_(b"branch merge")))
     if ctx.branch():
         edittext.append(hgprefix(_(b"branch '%s'") % ctx.branch()))
     if bookmarks.isactivewdirparent(repo):
         edittext.append(hgprefix(_(b"bookmark '%s'") % repo._activebookmark))
     edittext.extend([hgprefix(_(b"subrepo %s") % s) for s in subs])
     edittext.extend([hgprefix(_(b"added %s") % f) for f in added])
     edittext.extend([hgprefix(_(b"changed %s") % f) for f in modified])
     edittext.extend([hgprefix(_(b"removed %s") % f) for f in removed])
     if not added and not modified and not removed:
         edittext.append(hgprefix(_(b"no files changed")))
     edittext.append(b"")
 
     return b"\n".join(edittext)
 
 
 def commitstatus(repo, node, branch, bheads=None, tip=None, opts=None):
     if opts is None:
         opts = {}
     ctx = repo[node]
     parents = ctx.parents()
 
     if tip is not None and repo.changelog.tip() == tip:
         # avoid reporting something like "committed new head" when
         # recommitting old changesets, and issue a helpful warning
         # for most instances
         repo.ui.warn(_(b"warning: commit already existed in the repository!\n"))
     elif (
         not opts.get(b'amend')
         and bheads
         and node not in bheads
         and not any(
             p.node() in bheads and p.branch() == branch for p in parents
         )
     ):
         repo.ui.status(_(b'created new head\n'))
         # The message is not printed for initial roots. For the other
         # changesets, it is printed in the following situations:
         #
         # Par column: for the 2 parents with ...
         #   N: null or no parent
         #   B: parent is on another named branch
         #   C: parent is a regular non head changeset
         #   H: parent was a branch head of the current branch
         # Msg column: whether we print "created new head" message
         # In the following, it is assumed that there already exists some
         # initial branch heads of the current branch, otherwise nothing is
         # printed anyway.
         #
         # Par Msg Comment
         # N N  y  additional topo root
         #
         # B N  y  additional branch root
         # C N  y  additional topo head
         # H N  n  usual case
         #
         # B B  y  weird additional branch root
         # C B  y  branch merge
         # H B  n  merge with named branch
         #
         # C C  y  additional head from merge
         # C H  n  merge with a head
         #
         # H H  n  head merge: head count decreases
 
     if not opts.get(b'close_branch'):
         for r in parents:
             if r.closesbranch() and r.branch() == branch:
                 repo.ui.status(
                     _(b'reopening closed branch head %d\n') % r.rev()
                 )
 
     if repo.ui.debugflag:
         repo.ui.write(
             _(b'committed changeset %d:%s\n') % (ctx.rev(), ctx.hex())
         )
     elif repo.ui.verbose:
         repo.ui.write(_(b'committed changeset %d:%s\n') % (ctx.rev(), ctx))
 
 
 def postcommitstatus(repo, pats, opts):
     return repo.status(match=scmutil.match(repo[None], pats, opts))
 
 
 def revert(ui, repo, ctx, *pats, **opts):
     opts = pycompat.byteskwargs(opts)
     parent, p2 = repo.dirstate.parents()
     node = ctx.node()
 
     mf = ctx.manifest()
     if node == p2:
         parent = p2
 
     # need all matching names in dirstate and manifest of target rev,
     # so have to walk both. do not print errors if files exist in one
     # but not other. in both cases, filesets should be evaluated against
     # workingctx to get consistent result (issue4497). this means 'set:**'
     # cannot be used to select missing files from target rev.
 
     # `names` is a mapping for all elements in working copy and target revision
     # The mapping is in the form:
     #   <abs path in repo> -> (<path from CWD>, <exactly specified by matcher?>)
     names = {}
     uipathfn = scmutil.getuipathfn(repo, legacyrelativevalue=True)
 
     with repo.wlock():
         ## filling of the `names` mapping
         # walk dirstate to fill `names`
 
         interactive = opts.get(b'interactive', False)
         wctx = repo[None]
         m = scmutil.match(wctx, pats, opts)
 
         # we'll need this later
         targetsubs = sorted(s for s in wctx.substate if m(s))
 
         if not m.always():
             matcher = matchmod.badmatch(m, lambda x, y: False)
             for abs in wctx.walk(matcher):
                 names[abs] = m.exact(abs)
 
             # walk target manifest to fill `names`
 
             def badfn(path, msg):
                 if path in names:
                     return
                 if path in ctx.substate:
                     return
                 path_ = path + b'/'
                 for f in names:
                     if f.startswith(path_):
                         return
                 ui.warn(b"%s: %s\n" % (uipathfn(path), msg))
 
             for abs in ctx.walk(matchmod.badmatch(m, badfn)):
                 if abs not in names:
                     names[abs] = m.exact(abs)
 
             # Find status of all file in `names`.
             m = scmutil.matchfiles(repo, names)
 
             changes = repo.status(
                 node1=node, match=m, unknown=True, ignored=True, clean=True
             )
         else:
             changes = repo.status(node1=node, match=m)
             for kind in changes:
                 for abs in kind:
                     names[abs] = m.exact(abs)
 
             m = scmutil.matchfiles(repo, names)
 
         modified = set(changes.modified)
         added = set(changes.added)
         removed = set(changes.removed)
         _deleted = set(changes.deleted)
         unknown = set(changes.unknown)
         unknown.update(changes.ignored)
         clean = set(changes.clean)
         modadded = set()
 
         # We need to account for the state of the file in the dirstate,
         # even when we revert against something else than parent. This will
         # slightly alter the behavior of revert (doing back up or not, delete
         # or just forget etc).
         if parent == node:
             dsmodified = modified
             dsadded = added
             dsremoved = removed
             # store all local modifications, useful later for rename detection
             localchanges = dsmodified | dsadded
             modified, added, removed = set(), set(), set()
         else:
             changes = repo.status(node1=parent, match=m)
             dsmodified = set(changes.modified)
             dsadded = set(changes.added)
             dsremoved = set(changes.removed)
             # store all local modifications, useful later for rename detection
             localchanges = dsmodified | dsadded
 
             # only take into account for removes between wc and target
             clean |= dsremoved - removed
             dsremoved &= removed
             # distinct between dirstate remove and other
             removed -= dsremoved
 
             modadded = added & dsmodified
             added -= modadded
 
             # tell newly modified apart.
             dsmodified &= modified
             dsmodified |= modified & dsadded  # dirstate added may need backup
             modified -= dsmodified
 
             # We need to wait for some post-processing to update this set
             # before making the distinction. The dirstate will be used for
             # that purpose.
             dsadded = added
 
         # in case of merge, files that are actually added can be reported as
         # modified, we need to post process the result
         if p2 != repo.nullid:
             mergeadd = set(dsmodified)
             for path in dsmodified:
                 if path in mf:
                     mergeadd.remove(path)
             dsadded |= mergeadd
             dsmodified -= mergeadd
 
         # if f is a rename, update `names` to also revert the source
         for f in localchanges:
             src = repo.dirstate.copied(f)
             # XXX should we check for rename down to target node?
             if (
                 src
                 and src not in names
                 and repo.dirstate.get_entry(src).removed
             ):
                 dsremoved.add(src)
                 names[src] = True
 
         # determine the exact nature of the deleted changesets
         deladded = set(_deleted)
         for path in _deleted:
             if path in mf:
                 deladded.remove(path)
         deleted = _deleted - deladded
 
         # distinguish between file to forget and the other
         added = set()
         for abs in dsadded:
             if not repo.dirstate.get_entry(abs).added:
                 added.add(abs)
         dsadded -= added
 
         for abs in deladded:
             if repo.dirstate.get_entry(abs).added:
                 dsadded.add(abs)
         deladded -= dsadded
 
         # For files marked as removed, we check if an unknown file is present at
         # the same path. If a such file exists it may need to be backed up.
         # Making the distinction at this stage helps have simpler backup
         # logic.
         removunk = set()
         for abs in removed:
             target = repo.wjoin(abs)
             if os.path.lexists(target):
                 removunk.add(abs)
         removed -= removunk
 
         dsremovunk = set()
         for abs in dsremoved:
             target = repo.wjoin(abs)
             if os.path.lexists(target):
                 dsremovunk.add(abs)
         dsremoved -= dsremovunk
 
         # action to be actually performed by revert
         # (<list of file>, message>) tuple
         actions = {
             b'revert': ([], _(b'reverting %s\n')),
             b'add': ([], _(b'adding %s\n')),
             b'remove': ([], _(b'removing %s\n')),
             b'drop': ([], _(b'removing %s\n')),
             b'forget': ([], _(b'forgetting %s\n')),
             b'undelete': ([], _(b'undeleting %s\n')),
             b'noop': (None, _(b'no changes needed to %s\n')),
             b'unknown': (None, _(b'file not managed: %s\n')),
         }
 
         # "constant" that convey the backup strategy.
         # All set to `discard` if `no-backup` is set do avoid checking
         # no_backup lower in the code.
         # These values are ordered for comparison purposes
         backupinteractive = 3  # do backup if interactively modified
         backup = 2  # unconditionally do backup
         check = 1  # check if the existing file differs from target
         discard = 0  # never do backup
         if opts.get(b'no_backup'):
             backupinteractive = backup = check = discard
         if interactive:
             dsmodifiedbackup = backupinteractive
         else:
             dsmodifiedbackup = backup
         tobackup = set()
 
         backupanddel = actions[b'remove']
         if not opts.get(b'no_backup'):
             backupanddel = actions[b'drop']
 
         disptable = (
             # dispatch table:
             #   file state
             #   action
             #   make backup
             ## Sets that results that will change file on disk
             # Modified compared to target, no local change
             (modified, actions[b'revert'], discard),
             # Modified compared to target, but local file is deleted
             (deleted, actions[b'revert'], discard),
             # Modified compared to target, local change
             (dsmodified, actions[b'revert'], dsmodifiedbackup),
             # Added since target
             (added, actions[b'remove'], discard),
             # Added in working directory
             (dsadded, actions[b'forget'], discard),
             # Added since target, have local modification
             (modadded, backupanddel, backup),
             # Added since target but file is missing in working directory
             (deladded, actions[b'drop'], discard),
             # Removed since  target, before working copy parent
             (removed, actions[b'add'], discard),
             # Same as `removed` but an unknown file exists at the same path
             (removunk, actions[b'add'], check),
             # Removed since targe, marked as such in working copy parent
             (dsremoved, actions[b'undelete'], discard),
             # Same as `dsremoved` but an unknown file exists at the same path
             (dsremovunk, actions[b'undelete'], check),
             ## the following sets does not result in any file changes
             # File with no modification
             (clean, actions[b'noop'], discard),
             # Existing file, not tracked anywhere
             (unknown, actions[b'unknown'], discard),
         )
 
         for abs, exact in sorted(names.items()):
             # target file to be touch on disk (relative to cwd)
             target = repo.wjoin(abs)
             # search the entry in the dispatch table.
             # if the file is in any of these sets, it was touched in the working
             # directory parent and we are sure it needs to be reverted.
             for table, (xlist, msg), dobackup in disptable:
                 if abs not in table:
                     continue
                 if xlist is not None:
                     xlist.append(abs)
                     if dobackup:
                         # If in interactive mode, don't automatically create
                         # .orig files (issue4793)
                         if dobackup == backupinteractive:
                             tobackup.add(abs)
                         elif backup <= dobackup or wctx[abs].cmp(ctx[abs]):
                             absbakname = scmutil.backuppath(ui, repo, abs)
                             bakname = os.path.relpath(
                                 absbakname, start=repo.root
                             )
                             ui.note(
                                 _(b'saving current version of %s as %s\n')
                                 % (uipathfn(abs), uipathfn(bakname))
                             )
                             if not opts.get(b'dry_run'):
                                 if interactive:
                                     util.copyfile(target, absbakname)
                                 else:
                                     util.rename(target, absbakname)
                     if opts.get(b'dry_run'):
                         if ui.verbose or not exact:
                             ui.status(msg % uipathfn(abs))
                 elif exact:
                     ui.warn(msg % uipathfn(abs))
                 break
 
         if not opts.get(b'dry_run'):
             needdata = (b'revert', b'add', b'undelete')
             oplist = [actions[name][0] for name in needdata]
             prefetch = scmutil.prefetchfiles
             matchfiles = scmutil.matchfiles(
                 repo, [f for sublist in oplist for f in sublist]
             )
             prefetch(
                 repo,
                 [(ctx.rev(), matchfiles)],
             )
             match = scmutil.match(repo[None], pats)
             _performrevert(
                 repo,
                 ctx,
                 names,
                 uipathfn,
                 actions,
                 match,
                 interactive,
                 tobackup,
             )
 
         if targetsubs:
             # Revert the subrepos on the revert list
             for sub in targetsubs:
                 try:
                     wctx.sub(sub).revert(
                         ctx.substate[sub], *pats, **pycompat.strkwargs(opts)
                     )
                 except KeyError:
                     raise error.Abort(
                         b"subrepository '%s' does not exist in %s!"
                         % (sub, short(ctx.node()))
                     )
 
 
 def _performrevert(
     repo,
     ctx,
     names,
     uipathfn,
     actions,
     match,
     interactive=False,
     tobackup=None,
 ):
     """function that actually perform all the actions computed for revert
 
     This is an independent function to let extension to plug in and react to
     the imminent revert.
 
     Make sure you have the working directory locked when calling this function.
     """
     parent, p2 = repo.dirstate.parents()
     node = ctx.node()
     excluded_files = []
 
     def checkout(f):
         fc = ctx[f]
         repo.wwrite(f, fc.data(), fc.flags())
 
     def doremove(f):
         try:
             rmdir = repo.ui.configbool(b'experimental', b'removeemptydirs')
             repo.wvfs.unlinkpath(f, rmdir=rmdir)
         except OSError:
             pass
         repo.dirstate.set_untracked(f)
 
     def prntstatusmsg(action, f):
         exact = names[f]
         if repo.ui.verbose or not exact:
             repo.ui.status(actions[action][1] % uipathfn(f))
 
     audit_path = pathutil.pathauditor(repo.root, cached=True)
     for f in actions[b'forget'][0]:
         if interactive:
             choice = repo.ui.promptchoice(
                 _(b"forget added file %s (Yn)?$$ &Yes $$ &No") % uipathfn(f)
             )
             if choice == 0:
                 prntstatusmsg(b'forget', f)
                 repo.dirstate.set_untracked(f)
             else:
                 excluded_files.append(f)
         else:
             prntstatusmsg(b'forget', f)
             repo.dirstate.set_untracked(f)
     for f in actions[b'remove'][0]:
         audit_path(f)
         if interactive:
             choice = repo.ui.promptchoice(
                 _(b"remove added file %s (Yn)?$$ &Yes $$ &No") % uipathfn(f)
             )
             if choice == 0:
                 prntstatusmsg(b'remove', f)
                 doremove(f)
             else:
                 excluded_files.append(f)
         else:
             prntstatusmsg(b'remove', f)
             doremove(f)
     for f in actions[b'drop'][0]:
         audit_path(f)
         prntstatusmsg(b'drop', f)
         repo.dirstate.set_untracked(f)
 
     # We are reverting to our parent. If possible, we had like `hg status`
     # to report the file as clean. We have to be less agressive for
     # merges to avoid losing information about copy introduced by the merge.
     # This might comes with bugs ?
     reset_copy = p2 == repo.nullid
 
     def normal(filename):
         return repo.dirstate.set_tracked(filename, reset_copy=reset_copy)
 
     newlyaddedandmodifiedfiles = set()
     if interactive:
         # Prompt the user for changes to revert
         torevert = [f for f in actions[b'revert'][0] if f not in excluded_files]
         m = scmutil.matchfiles(repo, torevert)
         diffopts = patch.difffeatureopts(
             repo.ui,
             whitespace=True,
             section=b'commands',
             configprefix=b'revert.interactive.',
         )
         diffopts.nodates = True
         diffopts.git = True
         operation = b'apply'
         if node == parent:
             if repo.ui.configbool(
                 b'experimental', b'revert.interactive.select-to-keep'
             ):
                 operation = b'keep'
             else:
                 operation = b'discard'
 
         if operation == b'apply':
             diff = patch.diff(repo, None, ctx.node(), m, opts=diffopts)
         else:
             diff = patch.diff(repo, ctx.node(), None, m, opts=diffopts)
         original_headers = patch.parsepatch(diff)
 
         try:
 
             chunks, opts = recordfilter(
                 repo.ui, original_headers, match, operation=operation
             )
             if operation == b'discard':
                 chunks = patch.reversehunks(chunks)
 
         except error.PatchParseError as err:
             raise error.InputError(_(b'error parsing patch: %s') % err)
         except error.PatchApplicationError as err:
             raise error.StateError(_(b'error applying patch: %s') % err)
 
         # FIXME: when doing an interactive revert of a copy, there's no way of
         # performing a partial revert of the added file, the only option is
         # "remove added file <name> (Yn)?", so we don't need to worry about the
         # alsorestore value. Ideally we'd be able to partially revert
         # copied/renamed files.
         newlyaddedandmodifiedfiles, unusedalsorestore = newandmodified(chunks)
         if tobackup is None:
             tobackup = set()
         # Apply changes
         fp = stringio()
         # chunks are serialized per file, but files aren't sorted
         for f in sorted({c.header.filename() for c in chunks if ishunk(c)}):
             prntstatusmsg(b'revert', f)
         files = set()
         for c in chunks:
             if ishunk(c):
                 abs = c.header.filename()
                 # Create a backup file only if this hunk should be backed up
                 if c.header.filename() in tobackup:
                     target = repo.wjoin(abs)
                     bakname = scmutil.backuppath(repo.ui, repo, abs)
                     util.copyfile(target, bakname)
                     tobackup.remove(abs)
                 if abs not in files:
                     files.add(abs)
                     if operation == b'keep':
                         checkout(abs)
             c.write(fp)
         dopatch = fp.tell()
         fp.seek(0)
         if dopatch:
             try:
                 patch.internalpatch(repo.ui, repo, fp, 1, eolmode=None)
             except error.PatchParseError as err:
                 raise error.InputError(pycompat.bytestr(err))
             except error.PatchApplicationError as err:
                 raise error.StateError(pycompat.bytestr(err))
         del fp
     else:
         for f in actions[b'revert'][0]:
             prntstatusmsg(b'revert', f)
             checkout(f)
             if normal:
                 normal(f)
 
     for f in actions[b'add'][0]:
         # Don't checkout modified files, they are already created by the diff
         if f in newlyaddedandmodifiedfiles:
             continue
 
         if interactive:
             choice = repo.ui.promptchoice(
                 _(b"add new file %s (Yn)?$$ &Yes $$ &No") % uipathfn(f)
             )
             if choice != 0:
                 continue
         prntstatusmsg(b'add', f)
         checkout(f)
         repo.dirstate.set_tracked(f)
 
     for f in actions[b'undelete'][0]:
         if interactive:
             choice = repo.ui.promptchoice(
                 _(b"add back removed file %s (Yn)?$$ &Yes $$ &No") % f
             )
             if choice == 0:
                 prntstatusmsg(b'undelete', f)
                 checkout(f)
                 normal(f)
             else:
                 excluded_files.append(f)
         else:
             prntstatusmsg(b'undelete', f)
             checkout(f)
             normal(f)
 
     copied = copies.pathcopies(repo[parent], ctx)
 
     for f in (
         actions[b'add'][0] + actions[b'undelete'][0] + actions[b'revert'][0]
     ):
         if f in copied:
             repo.dirstate.copy(copied[f], f)
 
 
 # a list of (ui, repo, otherpeer, opts, missing) functions called by
 # commands.outgoing.  "missing" is "missing" of the result of
 # "findcommonoutgoing()"
 outgoinghooks = util.hooks()
 
 # a list of (ui, repo) functions called by commands.summary
 summaryhooks = util.hooks()
 
 # a list of (ui, repo, opts, changes) functions called by commands.summary.
 #
 # functions should return tuple of booleans below, if 'changes' is None:
 #  (whether-incomings-are-needed, whether-outgoings-are-needed)
 #
 # otherwise, 'changes' is a tuple of tuples below:
 #  - (sourceurl, sourcebranch, sourcepeer, incoming)
 #  - (desturl,   destbranch,   destpeer,   outgoing)
 summaryremotehooks = util.hooks()
 
 
 def checkunfinished(repo, commit=False, skipmerge=False):
     """Look for an unfinished multistep operation, like graft, and abort
     if found. It's probably good to check this right before
     bailifchanged().
     """
     # Check for non-clearable states first, so things like rebase will take
     # precedence over update.
     for state in statemod._unfinishedstates:
         if (
             state._clearable
             or (commit and state._allowcommit)
             or state._reportonly
         ):
             continue
         if state.isunfinished(repo):
             raise error.StateError(state.msg(), hint=state.hint())
 
     for s in statemod._unfinishedstates:
         if (
             not s._clearable
             or (commit and s._allowcommit)
             or (s._opname == b'merge' and skipmerge)
             or s._reportonly
         ):
             continue
         if s.isunfinished(repo):
             raise error.StateError(s.msg(), hint=s.hint())
 
 
 def clearunfinished(repo):
     """Check for unfinished operations (as above), and clear the ones
     that are clearable.
     """
     for state in statemod._unfinishedstates:
         if state._reportonly:
             continue
         if not state._clearable and state.isunfinished(repo):
             raise error.StateError(state.msg(), hint=state.hint())
 
     for s in statemod._unfinishedstates:
         if s._opname == b'merge' or s._reportonly:
             continue
         if s._clearable and s.isunfinished(repo):
             util.unlink(repo.vfs.join(s._fname))
 
 
 def getunfinishedstate(repo):
     """Checks for unfinished operations and returns statecheck object
     for it"""
     for state in statemod._unfinishedstates:
         if state.isunfinished(repo):
             return state
     return None
 
 
 def howtocontinue(repo):
     """Check for an unfinished operation and return the command to finish
     it.
 
     statemod._unfinishedstates list is checked for an unfinished operation
     and the corresponding message to finish it is generated if a method to
     continue is supported by the operation.
 
     Returns a (msg, warning) tuple. 'msg' is a string and 'warning' is
     a boolean.
     """
     contmsg = _(b"continue: %s")
     for state in statemod._unfinishedstates:
         if not state._continueflag:
             continue
         if state.isunfinished(repo):
             return contmsg % state.continuemsg(), True
     if repo[None].dirty(missing=True, merge=False, branch=False):
         return contmsg % _(b"hg commit"), False
     return None, None
 
 
 def checkafterresolved(repo):
     """Inform the user about the next action after completing hg resolve
 
     If there's a an unfinished operation that supports continue flag,
     howtocontinue will yield repo.ui.warn as the reporter.
 
     Otherwise, it will yield repo.ui.note.
     """
     msg, warning = howtocontinue(repo)
     if msg is not None:
         if warning:
             repo.ui.warn(b"%s\n" % msg)
         else:
             repo.ui.note(b"%s\n" % msg)
 
 
 def wrongtooltocontinue(repo, task):
     """Raise an abort suggesting how to properly continue if there is an
     active task.
 
     Uses howtocontinue() to find the active task.
 
     If there's no task (repo.ui.note for 'hg commit'), it does not offer
     a hint.
     """
     after = howtocontinue(repo)
     hint = None
     if after[1]:
         hint = after[0]
     raise error.StateError(_(b'no %s in progress') % task, hint=hint)
 
 
 def abortgraft(ui, repo, graftstate):
     """abort the interrupted graft and rollbacks to the state before interrupted
     graft"""
     if not graftstate.exists():
         raise error.StateError(_(b"no interrupted graft to abort"))
     statedata = readgraftstate(repo, graftstate)
     newnodes = statedata.get(b'newnodes')
     if newnodes is None:
         # and old graft state which does not have all the data required to abort
         # the graft
         raise error.Abort(_(b"cannot abort using an old graftstate"))
 
     # changeset from which graft operation was started
     if len(newnodes) > 0:
         startctx = repo[newnodes[0]].p1()
     else:
         startctx = repo[b'.']
     # whether to strip or not
     cleanup = False
 
     if newnodes:
         newnodes = [repo[r].rev() for r in newnodes]
         cleanup = True
         # checking that none of the newnodes turned public or is public
         immutable = [c for c in newnodes if not repo[c].mutable()]
         if immutable:
             repo.ui.warn(
                 _(b"cannot clean up public changesets %s\n")
                 % b', '.join(bytes(repo[r]) for r in immutable),
                 hint=_(b"see 'hg help phases' for details"),
             )
             cleanup = False
 
         # checking that no new nodes are created on top of grafted revs
         desc = set(repo.changelog.descendants(newnodes))
         if desc - set(newnodes):
             repo.ui.warn(
                 _(
                     b"new changesets detected on destination "
                     b"branch, can't strip\n"
                 )
             )
             cleanup = False
 
         if cleanup:
             with repo.wlock(), repo.lock():
                 mergemod.clean_update(startctx)
                 # stripping the new nodes created
                 strippoints = [
                     c.node() for c in repo.set(b"roots(%ld)", newnodes)
                 ]
                 repair.strip(repo.ui, repo, strippoints, backup=False)
 
     if not cleanup:
         # we don't update to the startnode if we can't strip
         startctx = repo[b'.']
         mergemod.clean_update(startctx)
 
     ui.status(_(b"graft aborted\n"))
     ui.status(_(b"working directory is now at %s\n") % startctx.hex()[:12])
     graftstate.delete()
     return 0
 
 
 def readgraftstate(repo, graftstate):
     # type: (Any, statemod.cmdstate) -> Dict[bytes, Any]
     """read the graft state file and return a dict of the data stored in it"""
     try:
         return graftstate.read()
     except error.CorruptedState:
         nodes = repo.vfs.read(b'graftstate').splitlines()
         return {b'nodes': nodes}
 
 
 def hgabortgraft(ui, repo):
     """abort logic for aborting graft using 'hg abort'"""
     with repo.wlock():
         graftstate = statemod.cmdstate(repo, b'graftstate')
         return abortgraft(ui, repo, graftstate)

mercurial/scmutil.py

@@ -1,2313 +1,2315 b''
 # scmutil.py - Mercurial core utility functions
 #
 #  Copyright Olivia Mackall <[e-mail]>
 #
 # This software may be used and distributed according to the terms of the
 # GNU General Public License version 2 or any later version.
 
 
 import binascii
 import errno
 import glob
 import os
 import posixpath
 import re
 import subprocess
 import weakref
 
 from .i18n import _
 from .node import (
     bin,
     hex,
     nullrev,
     short,
     wdirrev,
 )
 from .pycompat import getattr
 from .thirdparty import attr
 from . import (
     copies as copiesmod,
     encoding,
     error,
     match as matchmod,
     obsolete,
     obsutil,
     pathutil,
     phases,
     policy,
     pycompat,
     requirements as requirementsmod,
     revsetlang,
     similar,
     smartset,
     url,
     util,
     vfs,
 )
 
 from .utils import (
     hashutil,
     procutil,
     stringutil,
 )
 
 if pycompat.iswindows:
     from . import scmwindows as scmplatform
 else:
     from . import scmposix as scmplatform
 
 parsers = policy.importmod('parsers')
 rustrevlog = policy.importrust('revlog')
 
 termsize = scmplatform.termsize
 
 
 @attr.s(slots=True, repr=False)
 class status:
     """Struct with a list of files per status.
 
     The 'deleted', 'unknown' and 'ignored' properties are only
     relevant to the working copy.
     """
 
     modified = attr.ib(default=attr.Factory(list))
     added = attr.ib(default=attr.Factory(list))
     removed = attr.ib(default=attr.Factory(list))
     deleted = attr.ib(default=attr.Factory(list))
     unknown = attr.ib(default=attr.Factory(list))
     ignored = attr.ib(default=attr.Factory(list))
     clean = attr.ib(default=attr.Factory(list))
 
     def __iter__(self):
         yield self.modified
         yield self.added
         yield self.removed
         yield self.deleted
         yield self.unknown
         yield self.ignored
         yield self.clean
 
     def __repr__(self):
         return (
             r'<status modified=%s, added=%s, removed=%s, deleted=%s, '
             r'unknown=%s, ignored=%s, clean=%s>'
         ) % tuple(pycompat.sysstr(stringutil.pprint(v)) for v in self)
 
 
 def itersubrepos(ctx1, ctx2):
     """find subrepos in ctx1 or ctx2"""
     # Create a (subpath, ctx) mapping where we prefer subpaths from
     # ctx1. The subpaths from ctx2 are important when the .hgsub file
     # has been modified (in ctx2) but not yet committed (in ctx1).
     subpaths = dict.fromkeys(ctx2.substate, ctx2)
     subpaths.update(dict.fromkeys(ctx1.substate, ctx1))
 
     missing = set()
 
     for subpath in ctx2.substate:
         if subpath not in ctx1.substate:
             del subpaths[subpath]
             missing.add(subpath)
 
     for subpath, ctx in sorted(subpaths.items()):
         yield subpath, ctx.sub(subpath)
 
     # Yield an empty subrepo based on ctx1 for anything only in ctx2.  That way,
     # status and diff will have an accurate result when it does
     # 'sub.{status|diff}(rev2)'.  Otherwise, the ctx2 subrepo is compared
     # against itself.
     for subpath in missing:
         yield subpath, ctx2.nullsub(subpath, ctx1)
 
 
 def nochangesfound(ui, repo, excluded=None):
     """Report no changes for push/pull, excluded is None or a list of
     nodes excluded from the push/pull.
     """
     secretlist = []
     if excluded:
         for n in excluded:
             ctx = repo[n]
             if ctx.phase() >= phases.secret and not ctx.extinct():
                 secretlist.append(n)
 
     if secretlist:
         ui.status(
             _(b"no changes found (ignored %d secret changesets)\n")
             % len(secretlist)
         )
     else:
         ui.status(_(b"no changes found\n"))
 
 
 def callcatch(ui, func):
     """call func() with global exception handling
 
     return func() if no exception happens. otherwise do some error handling
     and return an exit code accordingly. does not handle all exceptions.
     """
     coarse_exit_code = -1
     detailed_exit_code = -1
     try:
         try:
             return func()
         except:  # re-raises
             ui.traceback()
             raise
     # Global exception handling, alphabetically
     # Mercurial-specific first, followed by built-in and library exceptions
     except error.LockHeld as inst:
         detailed_exit_code = 20
         if inst.errno == errno.ETIMEDOUT:
             reason = _(b'timed out waiting for lock held by %r') % (
                 pycompat.bytestr(inst.locker)
             )
         else:
             reason = _(b'lock held by %r') % inst.locker
         ui.error(
             _(b"abort: %s: %s\n")
             % (inst.desc or stringutil.forcebytestr(inst.filename), reason)
         )
         if not inst.locker:
             ui.error(_(b"(lock might be very busy)\n"))
     except error.LockUnavailable as inst:
         detailed_exit_code = 20
         ui.error(
             _(b"abort: could not lock %s: %s\n")
             % (
                 inst.desc or stringutil.forcebytestr(inst.filename),
                 encoding.strtolocal(inst.strerror),
             )
         )
     except error.RepoError as inst:
         if isinstance(inst, error.RepoLookupError):
             detailed_exit_code = 10
         ui.error(_(b"abort: %s\n") % inst)
         if inst.hint:
             ui.error(_(b"(%s)\n") % inst.hint)
     except error.ResponseError as inst:
         ui.error(_(b"abort: %s") % inst.args[0])
         msg = inst.args[1]
         if isinstance(msg, type(u'')):
             msg = pycompat.sysbytes(msg)
         if msg is None:
             ui.error(b"\n")
         elif not isinstance(msg, bytes):
             ui.error(b" %r\n" % (msg,))
         elif not msg:
             ui.error(_(b" empty string\n"))
         else:
             ui.error(b"\n%r\n" % pycompat.bytestr(stringutil.ellipsis(msg)))
     except error.CensoredNodeError as inst:
         ui.error(_(b"abort: file censored %s\n") % inst)
     except error.WdirUnsupported:
         ui.error(_(b"abort: working directory revision cannot be specified\n"))
     except error.Error as inst:
         if inst.detailed_exit_code is not None:
             detailed_exit_code = inst.detailed_exit_code
         if inst.coarse_exit_code is not None:
             coarse_exit_code = inst.coarse_exit_code
         ui.error(inst.format())
     except error.WorkerError as inst:
         # Don't print a message -- the worker already should have
         return inst.status_code
     except ImportError as inst:
         ui.error(_(b"abort: %s\n") % stringutil.forcebytestr(inst))
         m = stringutil.forcebytestr(inst).split()[-1]
         if m in b"mpatch bdiff".split():
             ui.error(_(b"(did you forget to compile extensions?)\n"))
         elif m in b"zlib".split():
             ui.error(_(b"(is your Python install correct?)\n"))
     except util.urlerr.httperror as inst:
         detailed_exit_code = 100
         ui.error(_(b"abort: %s\n") % stringutil.forcebytestr(inst))
     except util.urlerr.urlerror as inst:
         detailed_exit_code = 100
         try:  # usually it is in the form (errno, strerror)
             reason = inst.reason.args[1]
         except (AttributeError, IndexError):
             # it might be anything, for example a string
             reason = inst.reason
         if isinstance(reason, str):
             # SSLError of Python 2.7.9 contains a unicode
             reason = encoding.unitolocal(reason)
         ui.error(_(b"abort: error: %s\n") % stringutil.forcebytestr(reason))
     except (IOError, OSError) as inst:
         if (
             util.safehasattr(inst, b"args")
             and inst.args
             and inst.args[0] == errno.EPIPE
         ):
             pass
         elif getattr(inst, "strerror", None):  # common IOError or OSError
             if getattr(inst, "filename", None) is not None:
                 ui.error(
                     _(b"abort: %s: '%s'\n")
                     % (
                         encoding.strtolocal(inst.strerror),
                         stringutil.forcebytestr(inst.filename),
                     )
                 )
             else:
                 ui.error(_(b"abort: %s\n") % encoding.strtolocal(inst.strerror))
         else:  # suspicious IOError
             raise
     except MemoryError:
         ui.error(_(b"abort: out of memory\n"))
     except SystemExit as inst:
         # Commands shouldn't sys.exit directly, but give a return code.
         # Just in case catch this and and pass exit code to caller.
         detailed_exit_code = 254
         coarse_exit_code = inst.code
 
     if ui.configbool(b'ui', b'detailed-exit-code'):
         return detailed_exit_code
     else:
         return coarse_exit_code
 
 
 def checknewlabel(repo, lbl, kind):
     # Do not use the "kind" parameter in ui output.
     # It makes strings difficult to translate.
     if lbl in [b'tip', b'.', b'null']:
         raise error.InputError(_(b"the name '%s' is reserved") % lbl)
     for c in (b':', b'\0', b'\n', b'\r'):
         if c in lbl:
             raise error.InputError(
                 _(b"%r cannot be used in a name") % pycompat.bytestr(c)
             )
     try:
         int(lbl)
         if b'_' in lbl:
             # If label contains underscores, Python might consider it an
             # integer (with "_" as visual separators), but we do not.
             # See PEP 515 - Underscores in Numeric Literals.
             raise ValueError
         raise error.InputError(_(b"cannot use an integer as a name"))
     except ValueError:
         pass
     if lbl.strip() != lbl:
         raise error.InputError(
             _(b"leading or trailing whitespace in name %r") % lbl
         )
 
 
 def checkfilename(f):
     '''Check that the filename f is an acceptable filename for a tracked file'''
     if b'\r' in f or b'\n' in f:
         raise error.InputError(
             _(b"'\\n' and '\\r' disallowed in filenames: %r")
             % pycompat.bytestr(f)
         )
 
 
 def checkportable(ui, f):
     '''Check if filename f is portable and warn or abort depending on config'''
     checkfilename(f)
     abort, warn = checkportabilityalert(ui)
     if abort or warn:
         msg = util.checkwinfilename(f)
         if msg:
             msg = b"%s: %s" % (msg, procutil.shellquote(f))
             if abort:
                 raise error.InputError(msg)
             ui.warn(_(b"warning: %s\n") % msg)
 
 
 def checkportabilityalert(ui):
     """check if the user's config requests nothing, a warning, or abort for
     non-portable filenames"""
     val = ui.config(b'ui', b'portablefilenames')
     lval = val.lower()
     bval = stringutil.parsebool(val)
     abort = pycompat.iswindows or lval == b'abort'
     warn = bval or lval == b'warn'
     if bval is None and not (warn or abort or lval == b'ignore'):
         raise error.ConfigError(
             _(b"ui.portablefilenames value is invalid ('%s')") % val
         )
     return abort, warn
 
 
 class casecollisionauditor:
     def __init__(self, ui, abort, dirstate):
         self._ui = ui
         self._abort = abort
         allfiles = b'\0'.join(dirstate)
         self._loweredfiles = set(encoding.lower(allfiles).split(b'\0'))
         self._dirstate = dirstate
         # The purpose of _newfiles is so that we don't complain about
         # case collisions if someone were to call this object with the
         # same filename twice.
         self._newfiles = set()
 
     def __call__(self, f):
         if f in self._newfiles:
             return
         fl = encoding.lower(f)
         if fl in self._loweredfiles and f not in self._dirstate:
             msg = _(b'possible case-folding collision for %s') % f
             if self._abort:
                 raise error.StateError(msg)
             self._ui.warn(_(b"warning: %s\n") % msg)
         self._loweredfiles.add(fl)
         self._newfiles.add(f)
 
 
 def filteredhash(repo, maxrev, needobsolete=False):
     """build hash of filtered revisions in the current repoview.
 
     Multiple caches perform up-to-date validation by checking that the
     tiprev and tipnode stored in the cache file match the current repository.
     However, this is not sufficient for validating repoviews because the set
     of revisions in the view may change without the repository tiprev and
     tipnode changing.
 
     This function hashes all the revs filtered from the view (and, optionally,
     all obsolete revs) up to maxrev and returns that SHA-1 digest.
     """
     cl = repo.changelog
     if needobsolete:
         obsrevs = obsolete.getrevs(repo, b'obsolete')
         if not cl.filteredrevs and not obsrevs:
             return None
         key = (maxrev, hash(cl.filteredrevs), hash(obsrevs))
     else:
         if not cl.filteredrevs:
             return None
         key = maxrev
         obsrevs = frozenset()
 
     result = cl._filteredrevs_hashcache.get(key)
     if not result:
         revs = sorted(r for r in cl.filteredrevs | obsrevs if r <= maxrev)
         if revs:
             s = hashutil.sha1()
             for rev in revs:
                 s.update(b'%d;' % rev)
             result = s.digest()
             cl._filteredrevs_hashcache[key] = result
     return result
 
 
 def walkrepos(path, followsym=False, seen_dirs=None, recurse=False):
     """yield every hg repository under path, always recursively.
     The recurse flag will only control recursion into repo working dirs"""
 
     def errhandler(err):
         if err.filename == path:
             raise err
 
     samestat = getattr(os.path, 'samestat', None)
     if followsym and samestat is not None:
 
         def adddir(dirlst, dirname):
             dirstat = os.stat(dirname)
             match = any(samestat(dirstat, lstdirstat) for lstdirstat in dirlst)
             if not match:
                 dirlst.append(dirstat)
             return not match
 
     else:
         followsym = False
 
     if (seen_dirs is None) and followsym:
         seen_dirs = []
         adddir(seen_dirs, path)
     for root, dirs, files in os.walk(path, topdown=True, onerror=errhandler):
         dirs.sort()
         if b'.hg' in dirs:
             yield root  # found a repository
             qroot = os.path.join(root, b'.hg', b'patches')
             if os.path.isdir(os.path.join(qroot, b'.hg')):
                 yield qroot  # we have a patch queue repo here
             if recurse:
                 # avoid recursing inside the .hg directory
                 dirs.remove(b'.hg')
             else:
                 dirs[:] = []  # don't descend further
         elif followsym:
             newdirs = []
             for d in dirs:
                 fname = os.path.join(root, d)
                 if adddir(seen_dirs, fname):
                     if os.path.islink(fname):
                         for hgname in walkrepos(fname, True, seen_dirs):
                             yield hgname
                     else:
                         newdirs.append(d)
             dirs[:] = newdirs
 
 
 def binnode(ctx):
     """Return binary node id for a given basectx"""
     node = ctx.node()
     if node is None:
         return ctx.repo().nodeconstants.wdirid
     return node
 
 
 def intrev(ctx):
     """Return integer for a given basectx that can be used in comparison or
     arithmetic operation"""
     rev = ctx.rev()
     if rev is None:
         return wdirrev
     return rev
 
 
 def formatchangeid(ctx):
     """Format changectx as '{rev}:{node|formatnode}', which is the default
     template provided by logcmdutil.changesettemplater"""
     repo = ctx.repo()
     return formatrevnode(repo.ui, intrev(ctx), binnode(ctx))
 
 
 def formatrevnode(ui, rev, node):
     """Format given revision and node depending on the current verbosity"""
     if ui.debugflag:
         hexfunc = hex
     else:
         hexfunc = short
     return b'%d:%s' % (rev, hexfunc(node))
 
 
 def resolvehexnodeidprefix(repo, prefix):
     if prefix.startswith(b'x'):
         prefix = prefix[1:]
     try:
         # Uses unfiltered repo because it's faster when prefix is ambiguous/
         # This matches the shortesthexnodeidprefix() function below.
         node = repo.unfiltered().changelog._partialmatch(prefix)
     except error.AmbiguousPrefixLookupError:
         revset = repo.ui.config(
             b'experimental', b'revisions.disambiguatewithin'
         )
         if revset:
             # Clear config to avoid infinite recursion
             configoverrides = {
                 (b'experimental', b'revisions.disambiguatewithin'): None
             }
             with repo.ui.configoverride(configoverrides):
                 revs = repo.anyrevs([revset], user=True)
                 matches = []
                 for rev in revs:
                     node = repo.changelog.node(rev)
                     if hex(node).startswith(prefix):
                         matches.append(node)
                 if len(matches) == 1:
                     return matches[0]
         raise
     if node is None:
         return
     repo.changelog.rev(node)  # make sure node isn't filtered
     return node
 
 
 def mayberevnum(repo, prefix):
     """Checks if the given prefix may be mistaken for a revision number"""
     try:
         i = int(prefix)
         # if we are a pure int, then starting with zero will not be
         # confused as a rev; or, obviously, if the int is larger
         # than the value of the tip rev. We still need to disambiguate if
         # prefix == '0', since that *is* a valid revnum.
         if (prefix != b'0' and prefix[0:1] == b'0') or i >= len(repo):
             return False
         return True
     except ValueError:
         return False
 
 
 def shortesthexnodeidprefix(repo, node, minlength=1, cache=None):
     """Find the shortest unambiguous prefix that matches hexnode.
 
     If "cache" is not None, it must be a dictionary that can be used for
     caching between calls to this method.
     """
     # _partialmatch() of filtered changelog could take O(len(repo)) time,
     # which would be unacceptably slow. so we look for hash collision in
     # unfiltered space, which means some hashes may be slightly longer.
 
     minlength = max(minlength, 1)
 
     def disambiguate(prefix):
         """Disambiguate against revnums."""
         if repo.ui.configbool(b'experimental', b'revisions.prefixhexnode'):
             if mayberevnum(repo, prefix):
                 return b'x' + prefix
             else:
                 return prefix
 
         hexnode = hex(node)
         for length in range(len(prefix), len(hexnode) + 1):
             prefix = hexnode[:length]
             if not mayberevnum(repo, prefix):
                 return prefix
 
     cl = repo.unfiltered().changelog
     revset = repo.ui.config(b'experimental', b'revisions.disambiguatewithin')
     if revset:
         revs = None
         if cache is not None:
             revs = cache.get(b'disambiguationrevset')
         if revs is None:
             revs = repo.anyrevs([revset], user=True)
             if cache is not None:
                 cache[b'disambiguationrevset'] = revs
         if cl.rev(node) in revs:
             hexnode = hex(node)
             nodetree = None
             if cache is not None:
                 nodetree = cache.get(b'disambiguationnodetree')
             if not nodetree:
                 if util.safehasattr(parsers, 'nodetree'):
                     # The CExt is the only implementation to provide a nodetree
                     # class so far.
                     index = cl.index
                     if util.safehasattr(index, 'get_cindex'):
                         # the rust wrapped need to give access to its internal index
                         index = index.get_cindex()
                     nodetree = parsers.nodetree(index, len(revs))
                     for r in revs:
                         nodetree.insert(r)
                     if cache is not None:
                         cache[b'disambiguationnodetree'] = nodetree
             if nodetree is not None:
                 length = max(nodetree.shortest(node), minlength)
                 prefix = hexnode[:length]
                 return disambiguate(prefix)
             for length in range(minlength, len(hexnode) + 1):
                 matches = []
                 prefix = hexnode[:length]
                 for rev in revs:
                     otherhexnode = repo[rev].hex()
                     if prefix == otherhexnode[:length]:
                         matches.append(otherhexnode)
                 if len(matches) == 1:
                     return disambiguate(prefix)
 
     try:
         return disambiguate(cl.shortest(node, minlength))
     except error.LookupError:
         raise error.RepoLookupError()
 
 
 def isrevsymbol(repo, symbol):
     """Checks if a symbol exists in the repo.
 
     See revsymbol() for details. Raises error.AmbiguousPrefixLookupError if the
     symbol is an ambiguous nodeid prefix.
     """
     try:
         revsymbol(repo, symbol)
         return True
     except error.RepoLookupError:
         return False
 
 
 def revsymbol(repo, symbol):
     """Returns a context given a single revision symbol (as string).
 
     This is similar to revsingle(), but accepts only a single revision symbol,
     i.e. things like ".", "tip", "1234", "deadbeef", "my-bookmark" work, but
     not "max(public())".
     """
     if not isinstance(symbol, bytes):
         msg = (
             b"symbol (%s of type %s) was not a string, did you mean "
             b"repo[symbol]?" % (symbol, type(symbol))
         )
         raise error.ProgrammingError(msg)
     try:
         if symbol in (b'.', b'tip', b'null'):
             return repo[symbol]
 
         try:
             r = int(symbol)
             if b'%d' % r != symbol:
                 raise ValueError
             l = len(repo.changelog)
             if r < 0:
                 r += l
             if r < 0 or r >= l and r != wdirrev:
                 raise ValueError
             return repo[r]
         except error.FilteredIndexError:
             raise
         except (ValueError, OverflowError, IndexError):
             pass
 
         if len(symbol) == 2 * repo.nodeconstants.nodelen:
             try:
                 node = bin(symbol)
                 rev = repo.changelog.rev(node)
                 return repo[rev]
             except error.FilteredLookupError:
                 raise
             except (binascii.Error, LookupError):
                 pass
 
         # look up bookmarks through the name interface
         try:
             node = repo.names.singlenode(repo, symbol)
             rev = repo.changelog.rev(node)
             return repo[rev]
         except KeyError:
             pass
 
         node = resolvehexnodeidprefix(repo, symbol)
         if node is not None:
             rev = repo.changelog.rev(node)
             return repo[rev]
 
         raise error.RepoLookupError(_(b"unknown revision '%s'") % symbol)
 
     except error.WdirUnsupported:
         return repo[None]
     except (
         error.FilteredIndexError,
         error.FilteredLookupError,
         error.FilteredRepoLookupError,
     ):
         raise _filterederror(repo, symbol)
 
 
 def _filterederror(repo, changeid):
     """build an exception to be raised about a filtered changeid
 
     This is extracted in a function to help extensions (eg: evolve) to
     experiment with various message variants."""
     if repo.filtername.startswith(b'visible'):
 
         # Check if the changeset is obsolete
         unfilteredrepo = repo.unfiltered()
         ctx = revsymbol(unfilteredrepo, changeid)
 
         # If the changeset is obsolete, enrich the message with the reason
         # that made this changeset not visible
         if ctx.obsolete():
             msg = obsutil._getfilteredreason(repo, changeid, ctx)
         else:
             msg = _(b"hidden revision '%s'") % changeid
 
         hint = _(b'use --hidden to access hidden revisions')
 
         return error.FilteredRepoLookupError(msg, hint=hint)
     msg = _(b"filtered revision '%s' (not in '%s' subset)")
     msg %= (changeid, repo.filtername)
     return error.FilteredRepoLookupError(msg)
 
 
 def revsingle(repo, revspec, default=b'.', localalias=None):
     if not revspec and revspec != 0:
         return repo[default]
 
     l = revrange(repo, [revspec], localalias=localalias)
     if not l:
         raise error.InputError(_(b'empty revision set'))
     return repo[l.last()]
 
 
 def _pairspec(revspec):
     tree = revsetlang.parse(revspec)
     return tree and tree[0] in (
         b'range',
         b'rangepre',
         b'rangepost',
         b'rangeall',
     )
 
 
 def revpair(repo, revs):
     if not revs:
         return repo[b'.'], repo[None]
 
     l = revrange(repo, revs)
 
     if not l:
         raise error.InputError(_(b'empty revision range'))
 
     first = l.first()
     second = l.last()
 
     if (
         first == second
         and len(revs) >= 2
         and not all(revrange(repo, [r]) for r in revs)
     ):
         raise error.InputError(_(b'empty revision on one side of range'))
 
     # if top-level is range expression, the result must always be a pair
     if first == second and len(revs) == 1 and not _pairspec(revs[0]):
         return repo[first], repo[None]
 
     return repo[first], repo[second]
 
 
 def revrange(repo, specs, localalias=None):
     """Execute 1 to many revsets and return the union.
 
     This is the preferred mechanism for executing revsets using user-specified
     config options, such as revset aliases.
 
     The revsets specified by ``specs`` will be executed via a chained ``OR``
     expression. If ``specs`` is empty, an empty result is returned.
 
     ``specs`` can contain integers, in which case they are assumed to be
     revision numbers.
 
     It is assumed the revsets are already formatted. If you have arguments
     that need to be expanded in the revset, call ``revsetlang.formatspec()``
     and pass the result as an element of ``specs``.
 
     Specifying a single revset is allowed.
 
     Returns a ``smartset.abstractsmartset`` which is a list-like interface over
     integer revisions.
     """
     allspecs = []
     for spec in specs:
         if isinstance(spec, int):
             spec = revsetlang.formatspec(b'%d', spec)
         allspecs.append(spec)
     return repo.anyrevs(allspecs, user=True, localalias=localalias)
 
 
 def increasingwindows(windowsize=8, sizelimit=512):
     while True:
         yield windowsize
         if windowsize < sizelimit:
             windowsize *= 2
 
 
 def walkchangerevs(repo, revs, makefilematcher, prepare):
     """Iterate over files and the revs in a "windowed" way.
 
     Callers most commonly need to iterate backwards over the history
     in which they are interested. Doing so has awful (quadratic-looking)
     performance, so we use iterators in a "windowed" way.
 
     We walk a window of revisions in the desired order.  Within the
     window, we first walk forwards to gather data, then in the desired
     order (usually backwards) to display it.
 
     This function returns an iterator yielding contexts. Before
     yielding each context, the iterator will first call the prepare
     function on each context in the window in forward order."""
 
     if not revs:
         return []
     change = repo.__getitem__
 
     def iterate():
         it = iter(revs)
         stopiteration = False
         for windowsize in increasingwindows():
             nrevs = []
             for i in range(windowsize):
                 rev = next(it, None)
                 if rev is None:
                     stopiteration = True
                     break
                 nrevs.append(rev)
             for rev in sorted(nrevs):
                 ctx = change(rev)
                 prepare(ctx, makefilematcher(ctx))
             for rev in nrevs:
                 yield change(rev)
 
             if stopiteration:
                 break
 
     return iterate()
 
 
 def meaningfulparents(repo, ctx):
     """Return list of meaningful (or all if debug) parentrevs for rev.
 
     For merges (two non-nullrev revisions) both parents are meaningful.
     Otherwise the first parent revision is considered meaningful if it
     is not the preceding revision.
     """
     parents = ctx.parents()
     if len(parents) > 1:
         return parents
     if repo.ui.debugflag:
         return [parents[0], repo[nullrev]]
     if parents[0].rev() >= intrev(ctx) - 1:
         return []
     return parents
 
 
 def getuipathfn(repo, legacyrelativevalue=False, forcerelativevalue=None):
     """Return a function that produced paths for presenting to the user.
 
     The returned function takes a repo-relative path and produces a path
     that can be presented in the UI.
 
     Depending on the value of ui.relative-paths, either a repo-relative or
     cwd-relative path will be produced.
 
     legacyrelativevalue is the value to use if ui.relative-paths=legacy
 
     If forcerelativevalue is not None, then that value will be used regardless
     of what ui.relative-paths is set to.
     """
     if forcerelativevalue is not None:
         relative = forcerelativevalue
     else:
         config = repo.ui.config(b'ui', b'relative-paths')
         if config == b'legacy':
             relative = legacyrelativevalue
         else:
             relative = stringutil.parsebool(config)
             if relative is None:
                 raise error.ConfigError(
                     _(b"ui.relative-paths is not a boolean ('%s')") % config
                 )
 
     if relative:
         cwd = repo.getcwd()
         if cwd != b'':
             # this branch would work even if cwd == b'' (ie cwd = repo
             # root), but its generality makes the returned function slower
             pathto = repo.pathto
             return lambda f: pathto(f, cwd)
     if repo.ui.configbool(b'ui', b'slash'):
         return lambda f: f
     else:
         return util.localpath
 
 
 def subdiruipathfn(subpath, uipathfn):
     '''Create a new uipathfn that treats the file as relative to subpath.'''
     return lambda f: uipathfn(posixpath.join(subpath, f))
 
 
 def anypats(pats, opts):
     """Checks if any patterns, including --include and --exclude were given.
 
     Some commands (e.g. addremove) use this condition for deciding whether to
     print absolute or relative paths.
     """
     return bool(pats or opts.get(b'include') or opts.get(b'exclude'))
 
 
 def expandpats(pats):
     """Expand bare globs when running on windows.
     On posix we assume it already has already been done by sh."""
     if not util.expandglobs:
         return list(pats)
     ret = []
     for kindpat in pats:
         kind, pat = matchmod._patsplit(kindpat, None)
         if kind is None:
             try:
                 globbed = glob.glob(pat)
             except re.error:
                 globbed = [pat]
             if globbed:
                 ret.extend(globbed)
                 continue
         ret.append(kindpat)
     return ret
 
 
 def matchandpats(
     ctx, pats=(), opts=None, globbed=False, default=b'relpath', badfn=None
 ):
     """Return a matcher and the patterns that were used.
     The matcher will warn about bad matches, unless an alternate badfn callback
     is provided."""
     if opts is None:
         opts = {}
     if not globbed and default == b'relpath':
         pats = expandpats(pats or [])
 
     uipathfn = getuipathfn(ctx.repo(), legacyrelativevalue=True)
 
     def bad(f, msg):
         ctx.repo().ui.warn(b"%s: %s\n" % (uipathfn(f), msg))
 
     if badfn is None:
         badfn = bad
 
     m = ctx.match(
         pats,
         opts.get(b'include'),
         opts.get(b'exclude'),
         default,
         listsubrepos=opts.get(b'subrepos'),
         badfn=badfn,
     )
 
     if m.always():
         pats = []
     return m, pats
 
 
 def match(
     ctx, pats=(), opts=None, globbed=False, default=b'relpath', badfn=None
 ):
     '''Return a matcher that will warn about bad matches.'''
     return matchandpats(ctx, pats, opts, globbed, default, badfn=badfn)[0]
 
 
 def matchall(repo):
     '''Return a matcher that will efficiently match everything.'''
     return matchmod.always()
 
 
 def matchfiles(repo, files, badfn=None):
     '''Return a matcher that will efficiently match exactly these files.'''
     return matchmod.exact(files, badfn=badfn)
 
 
 def parsefollowlinespattern(repo, rev, pat, msg):
     """Return a file name from `pat` pattern suitable for usage in followlines
     logic.
     """
     if not matchmod.patkind(pat):
         return pathutil.canonpath(repo.root, repo.getcwd(), pat)
     else:
         ctx = repo[rev]
         m = matchmod.match(repo.root, repo.getcwd(), [pat], ctx=ctx)
         files = [f for f in ctx if m(f)]
         if len(files) != 1:
             raise error.ParseError(msg)
         return files[0]
 
 
 def getorigvfs(ui, repo):
     """return a vfs suitable to save 'orig' file
 
     return None if no special directory is configured"""
     origbackuppath = ui.config(b'ui', b'origbackuppath')
     if not origbackuppath:
         return None
     return vfs.vfs(repo.wvfs.join(origbackuppath))
 
 
 def backuppath(ui, repo, filepath):
     """customize where working copy backup files (.orig files) are created
 
     Fetch user defined path from config file: [ui] origbackuppath = <path>
     Fall back to default (filepath with .orig suffix) if not specified
 
     filepath is repo-relative
 
     Returns an absolute path
     """
     origvfs = getorigvfs(ui, repo)
     if origvfs is None:
         return repo.wjoin(filepath + b".orig")
 
     origbackupdir = origvfs.dirname(filepath)
     if not origvfs.isdir(origbackupdir) or origvfs.islink(origbackupdir):
         ui.note(_(b'creating directory: %s\n') % origvfs.join(origbackupdir))
 
         # Remove any files that conflict with the backup file's path
         for f in reversed(list(pathutil.finddirs(filepath))):
             if origvfs.isfileorlink(f):
                 ui.note(_(b'removing conflicting file: %s\n') % origvfs.join(f))
                 origvfs.unlink(f)
                 break
 
         origvfs.makedirs(origbackupdir)
 
     if origvfs.isdir(filepath) and not origvfs.islink(filepath):
         ui.note(
             _(b'removing conflicting directory: %s\n') % origvfs.join(filepath)
         )
         origvfs.rmtree(filepath, forcibly=True)
 
     return origvfs.join(filepath)
 
 
 class _containsnode:
     """proxy __contains__(node) to container.__contains__ which accepts revs"""
 
     def __init__(self, repo, revcontainer):
         self._torev = repo.changelog.rev
         self._revcontains = revcontainer.__contains__
 
     def __contains__(self, node):
         return self._revcontains(self._torev(node))
 
 
 def cleanupnodes(
     repo,
     replacements,
     operation,
     moves=None,
     metadata=None,
     fixphase=False,
     targetphase=None,
     backup=True,
 ):
     """do common cleanups when old nodes are replaced by new nodes
 
     That includes writing obsmarkers or stripping nodes, and moving bookmarks.
     (we might also want to move working directory parent in the future)
 
     By default, bookmark moves are calculated automatically from 'replacements',
     but 'moves' can be used to override that. Also, 'moves' may include
     additional bookmark moves that should not have associated obsmarkers.
 
     replacements is {oldnode: [newnode]} or a iterable of nodes if they do not
     have replacements. operation is a string, like "rebase".
 
     metadata is dictionary containing metadata to be stored in obsmarker if
     obsolescence is enabled.
     """
     assert fixphase or targetphase is None
     if not replacements and not moves:
         return
 
     # translate mapping's other forms
     if not util.safehasattr(replacements, b'items'):
         replacements = {(n,): () for n in replacements}
     else:
         # upgrading non tuple "source" to tuple ones for BC
         repls = {}
         for key, value in replacements.items():
             if not isinstance(key, tuple):
                 key = (key,)
             repls[key] = value
         replacements = repls
 
     # Unfiltered repo is needed since nodes in replacements might be hidden.
     unfi = repo.unfiltered()
 
     # Calculate bookmark movements
     if moves is None:
         moves = {}
         for oldnodes, newnodes in replacements.items():
             for oldnode in oldnodes:
                 if oldnode in moves:
                     continue
                 if len(newnodes) > 1:
                     # usually a split, take the one with biggest rev number
                     newnode = next(unfi.set(b'max(%ln)', newnodes)).node()
                 elif len(newnodes) == 0:
                     # move bookmark backwards
                     allreplaced = []
                     for rep in replacements:
                         allreplaced.extend(rep)
                     roots = list(
                         unfi.set(b'max((::%n) - %ln)', oldnode, allreplaced)
                     )
                     if roots:
                         newnode = roots[0].node()
                     else:
                         newnode = repo.nullid
                 else:
                     newnode = newnodes[0]
                 moves[oldnode] = newnode
 
     allnewnodes = [n for ns in replacements.values() for n in ns]
     toretract = {}
     toadvance = {}
     if fixphase:
         precursors = {}
         for oldnodes, newnodes in replacements.items():
             for oldnode in oldnodes:
                 for newnode in newnodes:
                     precursors.setdefault(newnode, []).append(oldnode)
 
         allnewnodes.sort(key=lambda n: unfi[n].rev())
         newphases = {}
 
         def phase(ctx):
             return newphases.get(ctx.node(), ctx.phase())
 
         for newnode in allnewnodes:
             ctx = unfi[newnode]
             parentphase = max(phase(p) for p in ctx.parents())
             if targetphase is None:
                 oldphase = max(
                     unfi[oldnode].phase() for oldnode in precursors[newnode]
                 )
                 newphase = max(oldphase, parentphase)
             else:
                 newphase = max(targetphase, parentphase)
             newphases[newnode] = newphase
             if newphase > ctx.phase():
                 toretract.setdefault(newphase, []).append(newnode)
             elif newphase < ctx.phase():
                 toadvance.setdefault(newphase, []).append(newnode)
 
     with repo.transaction(b'cleanup') as tr:
         # Move bookmarks
         bmarks = repo._bookmarks
         bmarkchanges = []
         for oldnode, newnode in moves.items():
             oldbmarks = repo.nodebookmarks(oldnode)
             if not oldbmarks:
                 continue
             from . import bookmarks  # avoid import cycle
 
             repo.ui.debug(
                 b'moving bookmarks %r from %s to %s\n'
                 % (
                     pycompat.rapply(pycompat.maybebytestr, oldbmarks),
                     hex(oldnode),
                     hex(newnode),
                 )
             )
             # Delete divergent bookmarks being parents of related newnodes
             deleterevs = repo.revs(
                 b'parents(roots(%ln & (::%n))) - parents(%n)',
                 allnewnodes,
                 newnode,
                 oldnode,
             )
             deletenodes = _containsnode(repo, deleterevs)
             for name in oldbmarks:
                 bmarkchanges.append((name, newnode))
                 for b in bookmarks.divergent2delete(repo, deletenodes, name):
                     bmarkchanges.append((b, None))
 
         if bmarkchanges:
             bmarks.applychanges(repo, tr, bmarkchanges)
 
         for phase, nodes in toretract.items():
             phases.retractboundary(repo, tr, phase, nodes)
         for phase, nodes in toadvance.items():
             phases.advanceboundary(repo, tr, phase, nodes)
 
         mayusearchived = repo.ui.config(b'experimental', b'cleanup-as-archived')
         # Obsolete or strip nodes
         if obsolete.isenabled(repo, obsolete.createmarkersopt):
             # If a node is already obsoleted, and we want to obsolete it
             # without a successor, skip that obssolete request since it's
             # unnecessary. That's the "if s or not isobs(n)" check below.
             # Also sort the node in topology order, that might be useful for
             # some obsstore logic.
             # NOTE: the sorting might belong to createmarkers.
             torev = unfi.changelog.rev
             sortfunc = lambda ns: torev(ns[0][0])
             rels = []
             for ns, s in sorted(replacements.items(), key=sortfunc):
                 rel = (tuple(unfi[n] for n in ns), tuple(unfi[m] for m in s))
                 rels.append(rel)
             if rels:
                 obsolete.createmarkers(
                     repo, rels, operation=operation, metadata=metadata
                 )
         elif phases.supportarchived(repo) and mayusearchived:
             # this assume we do not have "unstable" nodes above the cleaned ones
             allreplaced = set()
             for ns in replacements.keys():
                 allreplaced.update(ns)
             if backup:
                 from . import repair  # avoid import cycle
 
                 node = min(allreplaced, key=repo.changelog.rev)
                 repair.backupbundle(
                     repo, allreplaced, allreplaced, node, operation
                 )
             phases.retractboundary(repo, tr, phases.archived, allreplaced)
         else:
             from . import repair  # avoid import cycle
 
             tostrip = list(n for ns in replacements for n in ns)
             if tostrip:
                 repair.delayedstrip(
                     repo.ui, repo, tostrip, operation, backup=backup
                 )
 
 
-def addremove(repo, matcher, prefix, uipathfn, opts=None):
+def addremove(repo, matcher, prefix, uipathfn, opts=None, open_tr=None):
     if opts is None:
         opts = {}
     m = matcher
     dry_run = opts.get(b'dry_run')
     try:
         similarity = float(opts.get(b'similarity') or 0)
     except ValueError:
         raise error.InputError(_(b'similarity must be a number'))
     if similarity < 0 or similarity > 100:
         raise error.InputError(_(b'similarity must be between 0 and 100'))
     similarity /= 100.0
 
     ret = 0
 
     wctx = repo[None]
     for subpath in sorted(wctx.substate):
         submatch = matchmod.subdirmatcher(subpath, m)
         if opts.get(b'subrepos') or m.exact(subpath) or any(submatch.files()):
             sub = wctx.sub(subpath)
             subprefix = repo.wvfs.reljoin(prefix, subpath)
             subuipathfn = subdiruipathfn(subpath, uipathfn)
             try:
                 if sub.addremove(submatch, subprefix, subuipathfn, opts):
                     ret = 1
             except error.LookupError:
                 repo.ui.status(
                     _(b"skipping missing subrepository: %s\n")
                     % uipathfn(subpath)
                 )
 
     rejected = []
 
     def badfn(f, msg):
         if f in m.files():
             m.bad(f, msg)
         rejected.append(f)
 
     badmatch = matchmod.badmatch(m, badfn)
     added, unknown, deleted, removed, forgotten = _interestingfiles(
         repo, badmatch
     )
 
     unknownset = set(unknown + forgotten)
     toprint = unknownset.copy()
     toprint.update(deleted)
     for abs in sorted(toprint):
         if repo.ui.verbose or not m.exact(abs):
             if abs in unknownset:
                 status = _(b'adding %s\n') % uipathfn(abs)
                 label = b'ui.addremove.added'
             else:
                 status = _(b'removing %s\n') % uipathfn(abs)
                 label = b'ui.addremove.removed'
             repo.ui.status(status, label=label)
 
     renames = _findrenames(
         repo, m, added + unknown, removed + deleted, similarity, uipathfn
     )
 
-    if not dry_run:
+    if not dry_run and (unknown or forgotten or deleted or renames):
+        if open_tr is not None:
+            open_tr()
         _markchanges(repo, unknown + forgotten, deleted, renames)
 
     for f in rejected:
         if f in m.files():
             return 1
     return ret
 
 
 def marktouched(repo, files, similarity=0.0):
     """Assert that files have somehow been operated upon. files are relative to
     the repo root."""
     m = matchfiles(repo, files, badfn=lambda x, y: rejected.append(x))
     rejected = []
 
     added, unknown, deleted, removed, forgotten = _interestingfiles(repo, m)
 
     if repo.ui.verbose:
         unknownset = set(unknown + forgotten)
         toprint = unknownset.copy()
         toprint.update(deleted)
         for abs in sorted(toprint):
             if abs in unknownset:
                 status = _(b'adding %s\n') % abs
             else:
                 status = _(b'removing %s\n') % abs
             repo.ui.status(status)
 
     # TODO: We should probably have the caller pass in uipathfn and apply it to
     # the messages above too. legacyrelativevalue=True is consistent with how
     # it used to work.
     uipathfn = getuipathfn(repo, legacyrelativevalue=True)
     renames = _findrenames(
         repo, m, added + unknown, removed + deleted, similarity, uipathfn
     )
 
     _markchanges(repo, unknown + forgotten, deleted, renames)
 
     for f in rejected:
         if f in m.files():
             return 1
     return 0
 
 
 def _interestingfiles(repo, matcher):
     """Walk dirstate with matcher, looking for files that addremove would care
     about.
 
     This is different from dirstate.status because it doesn't care about
     whether files are modified or clean."""
     added, unknown, deleted, removed, forgotten = [], [], [], [], []
     audit_path = pathutil.pathauditor(repo.root, cached=True)
 
     ctx = repo[None]
     dirstate = repo.dirstate
     matcher = repo.narrowmatch(matcher, includeexact=True)
     walkresults = dirstate.walk(
         matcher,
         subrepos=sorted(ctx.substate),
         unknown=True,
         ignored=False,
         full=False,
     )
     for abs, st in walkresults.items():
         entry = dirstate.get_entry(abs)
         if (not entry.any_tracked) and audit_path.check(abs):
             unknown.append(abs)
         elif (not entry.removed) and not st:
             deleted.append(abs)
         elif entry.removed and st:
             forgotten.append(abs)
         # for finding renames
         elif entry.removed and not st:
             removed.append(abs)
         elif entry.added:
             added.append(abs)
 
     return added, unknown, deleted, removed, forgotten
 
 
 def _findrenames(repo, matcher, added, removed, similarity, uipathfn):
     '''Find renames from removed files to added ones.'''
     renames = {}
     if similarity > 0:
         for old, new, score in similar.findrenames(
             repo, added, removed, similarity
         ):
             if (
                 repo.ui.verbose
                 or not matcher.exact(old)
                 or not matcher.exact(new)
             ):
                 repo.ui.status(
                     _(
                         b'recording removal of %s as rename to %s '
                         b'(%d%% similar)\n'
                     )
                     % (uipathfn(old), uipathfn(new), score * 100)
                 )
             renames[new] = old
     return renames
 
 
 def _markchanges(repo, unknown, deleted, renames):
     """Marks the files in unknown as added, the files in deleted as removed,
     and the files in renames as copied."""
     wctx = repo[None]
     with repo.wlock():
         wctx.forget(deleted)
         wctx.add(unknown)
         for new, old in renames.items():
             wctx.copy(old, new)
 
 
 def getrenamedfn(repo, endrev=None):
     if copiesmod.usechangesetcentricalgo(repo):
 
         def getrenamed(fn, rev):
             ctx = repo[rev]
             p1copies = ctx.p1copies()
             if fn in p1copies:
                 return p1copies[fn]
             p2copies = ctx.p2copies()
             if fn in p2copies:
                 return p2copies[fn]
             return None
 
         return getrenamed
 
     rcache = {}
     if endrev is None:
         endrev = len(repo)
 
     def getrenamed(fn, rev):
         """looks up all renames for a file (up to endrev) the first
         time the file is given. It indexes on the changerev and only
         parses the manifest if linkrev != changerev.
         Returns rename info for fn at changerev rev."""
         if fn not in rcache:
             rcache[fn] = {}
             fl = repo.file(fn)
             for i in fl:
                 lr = fl.linkrev(i)
                 renamed = fl.renamed(fl.node(i))
                 rcache[fn][lr] = renamed and renamed[0]
                 if lr >= endrev:
                     break
         if rev in rcache[fn]:
             return rcache[fn][rev]
 
         # If linkrev != rev (i.e. rev not found in rcache) fallback to
         # filectx logic.
         try:
             return repo[rev][fn].copysource()
         except error.LookupError:
             return None
 
     return getrenamed
 
 
 def getcopiesfn(repo, endrev=None):
     if copiesmod.usechangesetcentricalgo(repo):
 
         def copiesfn(ctx):
             if ctx.p2copies():
                 allcopies = ctx.p1copies().copy()
                 # There should be no overlap
                 allcopies.update(ctx.p2copies())
                 return sorted(allcopies.items())
             else:
                 return sorted(ctx.p1copies().items())
 
     else:
         getrenamed = getrenamedfn(repo, endrev)
 
         def copiesfn(ctx):
             copies = []
             for fn in ctx.files():
                 rename = getrenamed(fn, ctx.rev())
                 if rename:
                     copies.append((fn, rename))
             return copies
 
     return copiesfn
 
 
 def dirstatecopy(ui, repo, wctx, src, dst, dryrun=False, cwd=None):
     """Update the dirstate to reflect the intent of copying src to dst. For
     different reasons it might not end with dst being marked as copied from src.
     """
     origsrc = repo.dirstate.copied(src) or src
     if dst == origsrc:  # copying back a copy?
         entry = repo.dirstate.get_entry(dst)
         if (entry.added or not entry.tracked) and not dryrun:
             repo.dirstate.set_tracked(dst)
     else:
         if repo.dirstate.get_entry(origsrc).added and origsrc == src:
             if not ui.quiet:
                 ui.warn(
                     _(
                         b"%s has not been committed yet, so no copy "
                         b"data will be stored for %s.\n"
                     )
                     % (repo.pathto(origsrc, cwd), repo.pathto(dst, cwd))
                 )
             if not repo.dirstate.get_entry(dst).tracked and not dryrun:
                 wctx.add([dst])
         elif not dryrun:
             wctx.copy(origsrc, dst)
 
 
 def movedirstate(repo, newctx, match=None):
     """Move the dirstate to newctx and adjust it as necessary.
 
     A matcher can be provided as an optimization. It is probably a bug to pass
     a matcher that doesn't match all the differences between the parent of the
     working copy and newctx.
     """
     oldctx = repo[b'.']
     ds = repo.dirstate
     copies = dict(ds.copies())
     ds.setparents(newctx.node(), repo.nullid)
     s = newctx.status(oldctx, match=match)
 
     for f in s.modified:
         ds.update_file_p1(f, p1_tracked=True)
 
     for f in s.added:
         ds.update_file_p1(f, p1_tracked=False)
 
     for f in s.removed:
         ds.update_file_p1(f, p1_tracked=True)
 
     # Merge old parent and old working dir copies
     oldcopies = copiesmod.pathcopies(newctx, oldctx, match)
     oldcopies.update(copies)
     copies = {dst: oldcopies.get(src, src) for dst, src in oldcopies.items()}
     # Adjust the dirstate copies
     for dst, src in copies.items():
         if src not in newctx or dst in newctx or not ds.get_entry(dst).added:
             src = None
         ds.copy(src, dst)
     repo._quick_access_changeid_invalidate()
 
 
 def filterrequirements(requirements):
     """filters the requirements into two sets:
 
     wcreq: requirements which should be written in .hg/requires
     storereq: which should be written in .hg/store/requires
 
     Returns (wcreq, storereq)
     """
     if requirementsmod.SHARESAFE_REQUIREMENT in requirements:
         wc, store = set(), set()
         for r in requirements:
             if r in requirementsmod.WORKING_DIR_REQUIREMENTS:
                 wc.add(r)
             else:
                 store.add(r)
         return wc, store
     return requirements, None
 
 
 def istreemanifest(repo):
     """returns whether the repository is using treemanifest or not"""
     return requirementsmod.TREEMANIFEST_REQUIREMENT in repo.requirements
 
 
 def writereporequirements(repo, requirements=None):
     """writes requirements for the repo
 
     Requirements are written to .hg/requires and .hg/store/requires based
     on whether share-safe mode is enabled and which requirements are wdir
     requirements and which are store requirements
     """
     if requirements:
         repo.requirements = requirements
     wcreq, storereq = filterrequirements(repo.requirements)
     if wcreq is not None:
         writerequires(repo.vfs, wcreq)
     if storereq is not None:
         writerequires(repo.svfs, storereq)
     elif repo.ui.configbool(b'format', b'usestore'):
         # only remove store requires if we are using store
         repo.svfs.tryunlink(b'requires')
 
 
 def writerequires(opener, requirements):
     with opener(b'requires', b'w', atomictemp=True) as fp:
         for r in sorted(requirements):
             fp.write(b"%s\n" % r)
 
 
 class filecachesubentry:
     def __init__(self, path, stat):
         self.path = path
         self.cachestat = None
         self._cacheable = None
 
         if stat:
             self.cachestat = filecachesubentry.stat(self.path)
 
             if self.cachestat:
                 self._cacheable = self.cachestat.cacheable()
             else:
                 # None means we don't know yet
                 self._cacheable = None
 
     def refresh(self):
         if self.cacheable():
             self.cachestat = filecachesubentry.stat(self.path)
 
     def cacheable(self):
         if self._cacheable is not None:
             return self._cacheable
 
         # we don't know yet, assume it is for now
         return True
 
     def changed(self):
         # no point in going further if we can't cache it
         if not self.cacheable():
             return True
 
         newstat = filecachesubentry.stat(self.path)
 
         # we may not know if it's cacheable yet, check again now
         if newstat and self._cacheable is None:
             self._cacheable = newstat.cacheable()
 
             # check again
             if not self._cacheable:
                 return True
 
         if self.cachestat != newstat:
             self.cachestat = newstat
             return True
         else:
             return False
 
     @staticmethod
     def stat(path):
         try:
             return util.cachestat(path)
         except FileNotFoundError:
             pass
 
 
 class filecacheentry:
     def __init__(self, paths, stat=True):
         self._entries = []
         for path in paths:
             self._entries.append(filecachesubentry(path, stat))
 
     def changed(self):
         '''true if any entry has changed'''
         for entry in self._entries:
             if entry.changed():
                 return True
         return False
 
     def refresh(self):
         for entry in self._entries:
             entry.refresh()
 
 
 class filecache:
     """A property like decorator that tracks files under .hg/ for updates.
 
     On first access, the files defined as arguments are stat()ed and the
     results cached. The decorated function is called. The results are stashed
     away in a ``_filecache`` dict on the object whose method is decorated.
 
     On subsequent access, the cached result is used as it is set to the
     instance dictionary.
 
     On external property set/delete operations, the caller must update the
     corresponding _filecache entry appropriately. Use __class__.<attr>.set()
     instead of directly setting <attr>.
 
     When using the property API, the cached data is always used if available.
     No stat() is performed to check if the file has changed.
 
     Others can muck about with the state of the ``_filecache`` dict. e.g. they
     can populate an entry before the property's getter is called. In this case,
     entries in ``_filecache`` will be used during property operations,
     if available. If the underlying file changes, it is up to external callers
     to reflect this by e.g. calling ``delattr(obj, attr)`` to remove the cached
     method result as well as possibly calling ``del obj._filecache[attr]`` to
     remove the ``filecacheentry``.
     """
 
     def __init__(self, *paths):
         self.paths = paths
 
     def tracked_paths(self, obj):
         return [self.join(obj, path) for path in self.paths]
 
     def join(self, obj, fname):
         """Used to compute the runtime path of a cached file.
 
         Users should subclass filecache and provide their own version of this
         function to call the appropriate join function on 'obj' (an instance
         of the class that its member function was decorated).
         """
         raise NotImplementedError
 
     def __call__(self, func):
         self.func = func
         self.sname = func.__name__
         self.name = pycompat.sysbytes(self.sname)
         return self
 
     def __get__(self, obj, type=None):
         # if accessed on the class, return the descriptor itself.
         if obj is None:
             return self
 
         assert self.sname not in obj.__dict__
 
         entry = obj._filecache.get(self.name)
 
         if entry:
             if entry.changed():
                 entry.obj = self.func(obj)
         else:
             paths = self.tracked_paths(obj)
 
             # We stat -before- creating the object so our cache doesn't lie if
             # a writer modified between the time we read and stat
             entry = filecacheentry(paths, True)
             entry.obj = self.func(obj)
 
             obj._filecache[self.name] = entry
 
         obj.__dict__[self.sname] = entry.obj
         return entry.obj
 
     # don't implement __set__(), which would make __dict__ lookup as slow as
     # function call.
 
     def set(self, obj, value):
         if self.name not in obj._filecache:
             # we add an entry for the missing value because X in __dict__
             # implies X in _filecache
             paths = self.tracked_paths(obj)
             ce = filecacheentry(paths, False)
             obj._filecache[self.name] = ce
         else:
             ce = obj._filecache[self.name]
 
         ce.obj = value  # update cached copy
         obj.__dict__[self.sname] = value  # update copy returned by obj.x
 
 
 def extdatasource(repo, source):
     """Gather a map of rev -> value dict from the specified source
 
     A source spec is treated as a URL, with a special case shell: type
     for parsing the output from a shell command.
 
     The data is parsed as a series of newline-separated records where
     each record is a revision specifier optionally followed by a space
     and a freeform string value. If the revision is known locally, it
     is converted to a rev, otherwise the record is skipped.
 
     Note that both key and value are treated as UTF-8 and converted to
     the local encoding. This allows uniformity between local and
     remote data sources.
     """
 
     spec = repo.ui.config(b"extdata", source)
     if not spec:
         raise error.Abort(_(b"unknown extdata source '%s'") % source)
 
     data = {}
     src = proc = None
     try:
         if spec.startswith(b"shell:"):
             # external commands should be run relative to the repo root
             cmd = spec[6:]
             proc = subprocess.Popen(
                 procutil.tonativestr(cmd),
                 shell=True,
                 bufsize=-1,
                 close_fds=procutil.closefds,
                 stdout=subprocess.PIPE,
                 cwd=procutil.tonativestr(repo.root),
             )
             src = proc.stdout
         else:
             # treat as a URL or file
             src = url.open(repo.ui, spec)
         for l in src:
             if b" " in l:
                 k, v = l.strip().split(b" ", 1)
             else:
                 k, v = l.strip(), b""
 
             k = encoding.tolocal(k)
             try:
                 data[revsingle(repo, k).rev()] = encoding.tolocal(v)
             except (error.LookupError, error.RepoLookupError, error.InputError):
                 pass  # we ignore data for nodes that don't exist locally
     finally:
         if proc:
             try:
                 proc.communicate()
             except ValueError:
                 # This happens if we started iterating src and then
                 # get a parse error on a line. It should be safe to ignore.
                 pass
         if src:
             src.close()
     if proc and proc.returncode != 0:
         raise error.Abort(
             _(b"extdata command '%s' failed: %s")
             % (cmd, procutil.explainexit(proc.returncode))
         )
 
     return data
 
 
 class progress:
     def __init__(self, ui, updatebar, topic, unit=b"", total=None):
         self.ui = ui
         self.pos = 0
         self.topic = topic
         self.unit = unit
         self.total = total
         self.debug = ui.configbool(b'progress', b'debug')
         self._updatebar = updatebar
 
     def __enter__(self):
         return self
 
     def __exit__(self, exc_type, exc_value, exc_tb):
         self.complete()
 
     def update(self, pos, item=b"", total=None):
         assert pos is not None
         if total:
             self.total = total
         self.pos = pos
         self._updatebar(self.topic, self.pos, item, self.unit, self.total)
         if self.debug:
             self._printdebug(item)
 
     def increment(self, step=1, item=b"", total=None):
         self.update(self.pos + step, item, total)
 
     def complete(self):
         self.pos = None
         self.unit = b""
         self.total = None
         self._updatebar(self.topic, self.pos, b"", self.unit, self.total)
 
     def _printdebug(self, item):
         unit = b''
         if self.unit:
             unit = b' ' + self.unit
         if item:
             item = b' ' + item
 
         if self.total:
             pct = 100.0 * self.pos / self.total
             self.ui.debug(
                 b'%s:%s %d/%d%s (%4.2f%%)\n'
                 % (self.topic, item, self.pos, self.total, unit, pct)
             )
         else:
             self.ui.debug(b'%s:%s %d%s\n' % (self.topic, item, self.pos, unit))
 
 
 def gdinitconfig(ui):
     """helper function to know if a repo should be created as general delta"""
     # experimental config: format.generaldelta
     return ui.configbool(b'format', b'generaldelta') or ui.configbool(
         b'format', b'usegeneraldelta'
     )
 
 
 def gddeltaconfig(ui):
     """helper function to know if incoming deltas should be optimized
 
     The `format.generaldelta` config is an old form of the config that also
     implies that incoming delta-bases should be never be trusted. This function
     exists for this purpose.
     """
     # experimental config: format.generaldelta
     return ui.configbool(b'format', b'generaldelta')
 
 
 class simplekeyvaluefile:
     """A simple file with key=value lines
 
     Keys must be alphanumerics and start with a letter, values must not
     contain '\n' characters"""
 
     firstlinekey = b'__firstline'
 
     def __init__(self, vfs, path, keys=None):
         self.vfs = vfs
         self.path = path
 
     def read(self, firstlinenonkeyval=False):
         """Read the contents of a simple key-value file
 
         'firstlinenonkeyval' indicates whether the first line of file should
         be treated as a key-value pair or reuturned fully under the
         __firstline key."""
         lines = self.vfs.readlines(self.path)
         d = {}
         if firstlinenonkeyval:
             if not lines:
                 e = _(b"empty simplekeyvalue file")
                 raise error.CorruptedState(e)
             # we don't want to include '\n' in the __firstline
             d[self.firstlinekey] = lines[0][:-1]
             del lines[0]
 
         try:
             # the 'if line.strip()' part prevents us from failing on empty
             # lines which only contain '\n' therefore are not skipped
             # by 'if line'
             updatedict = dict(
                 line[:-1].split(b'=', 1) for line in lines if line.strip()
             )
             if self.firstlinekey in updatedict:
                 e = _(b"%r can't be used as a key")
                 raise error.CorruptedState(e % self.firstlinekey)
             d.update(updatedict)
         except ValueError as e:
             raise error.CorruptedState(stringutil.forcebytestr(e))
         return d
 
     def write(self, data, firstline=None):
         """Write key=>value mapping to a file
         data is a dict. Keys must be alphanumerical and start with a letter.
         Values must not contain newline characters.
 
         If 'firstline' is not None, it is written to file before
         everything else, as it is, not in a key=value form"""
         lines = []
         if firstline is not None:
             lines.append(b'%s\n' % firstline)
 
         for k, v in data.items():
             if k == self.firstlinekey:
                 e = b"key name '%s' is reserved" % self.firstlinekey
                 raise error.ProgrammingError(e)
             if not k[0:1].isalpha():
                 e = b"keys must start with a letter in a key-value file"
                 raise error.ProgrammingError(e)
             if not k.isalnum():
                 e = b"invalid key name in a simple key-value file"
                 raise error.ProgrammingError(e)
             if b'\n' in v:
                 e = b"invalid value in a simple key-value file"
                 raise error.ProgrammingError(e)
             lines.append(b"%s=%s\n" % (k, v))
         with self.vfs(self.path, mode=b'wb', atomictemp=True) as fp:
             fp.write(b''.join(lines))
 
 
 _reportobsoletedsource = [
     b'debugobsolete',
     b'pull',
     b'push',
     b'serve',
     b'unbundle',
 ]
 
 _reportnewcssource = [
     b'pull',
     b'unbundle',
 ]
 
 
 def prefetchfiles(repo, revmatches):
     """Invokes the registered file prefetch functions, allowing extensions to
     ensure the corresponding files are available locally, before the command
     uses them.
 
     Args:
       revmatches: a list of (revision, match) tuples to indicate the files to
       fetch at each revision. If any of the match elements is None, it matches
       all files.
     """
 
     def _matcher(m):
         if m:
             assert isinstance(m, matchmod.basematcher)
             # The command itself will complain about files that don't exist, so
             # don't duplicate the message.
             return matchmod.badmatch(m, lambda fn, msg: None)
         else:
             return matchall(repo)
 
     revbadmatches = [(rev, _matcher(match)) for (rev, match) in revmatches]
 
     fileprefetchhooks(repo, revbadmatches)
 
 
 # a list of (repo, revs, match) prefetch functions
 fileprefetchhooks = util.hooks()
 
 # A marker that tells the evolve extension to suppress its own reporting
 _reportstroubledchangesets = True
 
 
 def registersummarycallback(repo, otr, txnname=b'', as_validator=False):
     """register a callback to issue a summary after the transaction is closed
 
     If as_validator is true, then the callbacks are registered as transaction
     validators instead
     """
 
     def txmatch(sources):
         return any(txnname.startswith(source) for source in sources)
 
     categories = []
 
     def reportsummary(func):
         """decorator for report callbacks."""
         # The repoview life cycle is shorter than the one of the actual
         # underlying repository. So the filtered object can die before the
         # weakref is used leading to troubles. We keep a reference to the
         # unfiltered object and restore the filtering when retrieving the
         # repository through the weakref.
         filtername = repo.filtername
         reporef = weakref.ref(repo.unfiltered())
 
         def wrapped(tr):
             repo = reporef()
             if filtername:
                 assert repo is not None  # help pytype
                 repo = repo.filtered(filtername)
             func(repo, tr)
 
         newcat = b'%02i-txnreport' % len(categories)
         if as_validator:
             otr.addvalidator(newcat, wrapped)
         else:
             otr.addpostclose(newcat, wrapped)
         categories.append(newcat)
         return wrapped
 
     @reportsummary
     def reportchangegroup(repo, tr):
         cgchangesets = tr.changes.get(b'changegroup-count-changesets', 0)
         cgrevisions = tr.changes.get(b'changegroup-count-revisions', 0)
         cgfiles = tr.changes.get(b'changegroup-count-files', 0)
         cgheads = tr.changes.get(b'changegroup-count-heads', 0)
         if cgchangesets or cgrevisions or cgfiles:
             htext = b""
             if cgheads:
                 htext = _(b" (%+d heads)") % cgheads
             msg = _(b"added %d changesets with %d changes to %d files%s\n")
             if as_validator:
                 msg = _(b"adding %d changesets with %d changes to %d files%s\n")
             assert repo is not None  # help pytype
             repo.ui.status(msg % (cgchangesets, cgrevisions, cgfiles, htext))
 
     if txmatch(_reportobsoletedsource):
 
         @reportsummary
         def reportobsoleted(repo, tr):
             obsoleted = obsutil.getobsoleted(repo, tr)
             newmarkers = len(tr.changes.get(b'obsmarkers', ()))
             if newmarkers:
                 repo.ui.status(_(b'%i new obsolescence markers\n') % newmarkers)
             if obsoleted:
                 msg = _(b'obsoleted %i changesets\n')
                 if as_validator:
                     msg = _(b'obsoleting %i changesets\n')
                 repo.ui.status(msg % len(obsoleted))
 
     if obsolete.isenabled(
         repo, obsolete.createmarkersopt
     ) and repo.ui.configbool(
         b'experimental', b'evolution.report-instabilities'
     ):
         instabilitytypes = [
             (b'orphan', b'orphan'),
             (b'phase-divergent', b'phasedivergent'),
             (b'content-divergent', b'contentdivergent'),
         ]
 
         def getinstabilitycounts(repo):
             filtered = repo.changelog.filteredrevs
             counts = {}
             for instability, revset in instabilitytypes:
                 counts[instability] = len(
                     set(obsolete.getrevs(repo, revset)) - filtered
                 )
             return counts
 
         oldinstabilitycounts = getinstabilitycounts(repo)
 
         @reportsummary
         def reportnewinstabilities(repo, tr):
             newinstabilitycounts = getinstabilitycounts(repo)
             for instability, revset in instabilitytypes:
                 delta = (
                     newinstabilitycounts[instability]
                     - oldinstabilitycounts[instability]
                 )
                 msg = getinstabilitymessage(delta, instability)
                 if msg:
                     repo.ui.warn(msg)
 
     if txmatch(_reportnewcssource):
 
         @reportsummary
         def reportnewcs(repo, tr):
             """Report the range of new revisions pulled/unbundled."""
             origrepolen = tr.changes.get(b'origrepolen', len(repo))
             unfi = repo.unfiltered()
             if origrepolen >= len(unfi):
                 return
 
             # Compute the bounds of new visible revisions' range.
             revs = smartset.spanset(repo, start=origrepolen)
             if revs:
                 minrev, maxrev = repo[revs.min()], repo[revs.max()]
 
                 if minrev == maxrev:
                     revrange = minrev
                 else:
                     revrange = b'%s:%s' % (minrev, maxrev)
                 draft = len(repo.revs(b'%ld and draft()', revs))
                 secret = len(repo.revs(b'%ld and secret()', revs))
                 if not (draft or secret):
                     msg = _(b'new changesets %s\n') % revrange
                 elif draft and secret:
                     msg = _(b'new changesets %s (%d drafts, %d secrets)\n')
                     msg %= (revrange, draft, secret)
                 elif draft:
                     msg = _(b'new changesets %s (%d drafts)\n')
                     msg %= (revrange, draft)
                 elif secret:
                     msg = _(b'new changesets %s (%d secrets)\n')
                     msg %= (revrange, secret)
                 else:
                     errormsg = b'entered unreachable condition'
                     raise error.ProgrammingError(errormsg)
                 repo.ui.status(msg)
 
             # search new changesets directly pulled as obsolete
             duplicates = tr.changes.get(b'revduplicates', ())
             obsadded = unfi.revs(
                 b'(%d: + %ld) and obsolete()', origrepolen, duplicates
             )
             cl = repo.changelog
             extinctadded = [r for r in obsadded if r not in cl]
             if extinctadded:
                 # They are not just obsolete, but obsolete and invisible
                 # we call them "extinct" internally but the terms have not been
                 # exposed to users.
                 msg = b'(%d other changesets obsolete on arrival)\n'
                 repo.ui.status(msg % len(extinctadded))
 
         @reportsummary
         def reportphasechanges(repo, tr):
             """Report statistics of phase changes for changesets pre-existing
             pull/unbundle.
             """
             origrepolen = tr.changes.get(b'origrepolen', len(repo))
             published = []
             for revs, (old, new) in tr.changes.get(b'phases', []):
                 if new != phases.public:
                     continue
                 published.extend(rev for rev in revs if rev < origrepolen)
             if not published:
                 return
             msg = _(b'%d local changesets published\n')
             if as_validator:
                 msg = _(b'%d local changesets will be published\n')
             repo.ui.status(msg % len(published))
 
 
 def getinstabilitymessage(delta, instability):
     """function to return the message to show warning about new instabilities
 
     exists as a separate function so that extension can wrap to show more
     information like how to fix instabilities"""
     if delta > 0:
         return _(b'%i new %s changesets\n') % (delta, instability)
 
 
 def nodesummaries(repo, nodes, maxnumnodes=4):
     if len(nodes) <= maxnumnodes or repo.ui.verbose:
         return b' '.join(short(h) for h in nodes)
     first = b' '.join(short(h) for h in nodes[:maxnumnodes])
     return _(b"%s and %d others") % (first, len(nodes) - maxnumnodes)
 
 
 def enforcesinglehead(repo, tr, desc, accountclosed, filtername):
     """check that no named branch has multiple heads"""
     if desc in (b'strip', b'repair'):
         # skip the logic during strip
         return
     visible = repo.filtered(filtername)
     # possible improvement: we could restrict the check to affected branch
     bm = visible.branchmap()
     for name in bm:
         heads = bm.branchheads(name, closed=accountclosed)
         if len(heads) > 1:
             msg = _(b'rejecting multiple heads on branch "%s"')
             msg %= name
             hint = _(b'%d heads: %s')
             hint %= (len(heads), nodesummaries(repo, heads))
             raise error.Abort(msg, hint=hint)
 
 
 def wrapconvertsink(sink):
     """Allow extensions to wrap the sink returned by convcmd.convertsink()
     before it is used, whether or not the convert extension was formally loaded.
     """
     return sink
 
 
 def unhidehashlikerevs(repo, specs, hiddentype):
     """parse the user specs and unhide changesets whose hash or revision number
     is passed.
 
     hiddentype can be: 1) 'warn': warn while unhiding changesets
                        2) 'nowarn': don't warn while unhiding changesets
 
     returns a repo object with the required changesets unhidden
     """
     if not specs:
         return repo
 
     if not repo.filtername or not repo.ui.configbool(
         b'experimental', b'directaccess'
     ):
         return repo
 
     if repo.filtername not in (b'visible', b'visible-hidden'):
         return repo
 
     symbols = set()
     for spec in specs:
         try:
             tree = revsetlang.parse(spec)
         except error.ParseError:  # will be reported by scmutil.revrange()
             continue
 
         symbols.update(revsetlang.gethashlikesymbols(tree))
 
     if not symbols:
         return repo
 
     revs = _getrevsfromsymbols(repo, symbols)
 
     if not revs:
         return repo
 
     if hiddentype == b'warn':
         unfi = repo.unfiltered()
         revstr = b", ".join([pycompat.bytestr(unfi[l]) for l in revs])
         repo.ui.warn(
             _(
                 b"warning: accessing hidden changesets for write "
                 b"operation: %s\n"
             )
             % revstr
         )
 
     # we have to use new filtername to separate branch/tags cache until we can
     # disbale these cache when revisions are dynamically pinned.
     return repo.filtered(b'visible-hidden', revs)
 
 
 def _getrevsfromsymbols(repo, symbols):
     """parse the list of symbols and returns a set of revision numbers of hidden
     changesets present in symbols"""
     revs = set()
     unfi = repo.unfiltered()
     unficl = unfi.changelog
     cl = repo.changelog
     tiprev = len(unficl)
     allowrevnums = repo.ui.configbool(b'experimental', b'directaccess.revnums')
     for s in symbols:
         try:
             n = int(s)
             if n <= tiprev:
                 if not allowrevnums:
                     continue
                 else:
                     if n not in cl:
                         revs.add(n)
                     continue
         except ValueError:
             pass
 
         try:
             s = resolvehexnodeidprefix(unfi, s)
         except (error.LookupError, error.WdirUnsupported):
             s = None
 
         if s is not None:
             rev = unficl.rev(s)
             if rev not in cl:
                 revs.add(rev)
 
     return revs
 
 
 def bookmarkrevs(repo, mark):
     """Select revisions reachable by a given bookmark
 
     If the bookmarked revision isn't a head, an empty set will be returned.
     """
     return repo.revs(format_bookmark_revspec(mark))
 
 
 def format_bookmark_revspec(mark):
     """Build a revset expression to select revisions reachable by a given
     bookmark"""
     mark = b'literal:' + mark
     return revsetlang.formatspec(
         b"ancestors(bookmark(%s)) - "
         b"ancestors(head() and not bookmark(%s)) - "
         b"ancestors(bookmark() and not bookmark(%s))",
         mark,
         mark,
         mark,
     )

tests/test-fncache.t

@@ -1,537 +1,533 b''
 #require repofncache
 
 An extension which will set fncache chunksize to 1 byte to make sure that logic
 does not break
 
   $ cat > chunksize.py <<EOF
   > from mercurial import store
   > store.fncache_chunksize = 1
   > EOF
 
   $ cat >> $HGRCPATH <<EOF
   > [extensions]
   > chunksize = $TESTTMP/chunksize.py
   > EOF
 
 Init repo1:
 
   $ hg init repo1
   $ cd repo1
   $ echo "some text" > a
   $ hg add
   adding a
   $ hg ci -m first
   $ cat .hg/store/fncache | sort
   data/a.i
 
 Testing a.i/b:
 
   $ mkdir a.i
   $ echo "some other text" > a.i/b
   $ hg add
   adding a.i/b
   $ hg ci -m second
   $ cat .hg/store/fncache | sort
   data/a.i
   data/a.i.hg/b.i
 
 Testing a.i.hg/c:
 
   $ mkdir a.i.hg
   $ echo "yet another text" > a.i.hg/c
   $ hg add
   adding a.i.hg/c
   $ hg ci -m third
   $ cat .hg/store/fncache | sort
   data/a.i
   data/a.i.hg.hg/c.i
   data/a.i.hg/b.i
 
 Testing verify:
 
   $ hg verify -q
 
   $ rm .hg/store/fncache
 
   $ hg verify
   checking changesets
   checking manifests
   crosschecking files in changesets and manifests
   checking files
    warning: revlog 'data/a.i' not in fncache!
    warning: revlog 'data/a.i.hg/c.i' not in fncache!
    warning: revlog 'data/a.i/b.i' not in fncache!
   checking dirstate
   checked 3 changesets with 3 changes to 3 files
   3 warnings encountered!
   hint: run "hg debugrebuildfncache" to recover from corrupt fncache
 
 Follow the hint to make sure it works
 
   $ hg debugrebuildfncache
   adding data/a.i
   adding data/a.i.hg/c.i
   adding data/a.i/b.i
   3 items added, 0 removed from fncache
 
   $ hg verify -q
 
   $ cd ..
 
 Non store repo:
 
   $ hg --config format.usestore=False init foo
   $ cd foo
   $ mkdir tst.d
   $ echo foo > tst.d/foo
   $ hg ci -Amfoo
   adding tst.d/foo
   $ find .hg | sort
   .hg
   .hg/00changelog.i
   .hg/00manifest.i
   .hg/cache
   .hg/cache/branch2-served
   .hg/cache/rbc-names-v1
   .hg/cache/rbc-revs-v1
   .hg/data
   .hg/data/tst.d.hg
   .hg/data/tst.d.hg/foo.i
   .hg/dirstate
   .hg/fsmonitor.state (fsmonitor !)
   .hg/last-message.txt
   .hg/phaseroots
   .hg/requires
   .hg/undo
-  .hg/undo.backup.dirstate
   .hg/undo.backupfiles
   .hg/undo.bookmarks
   .hg/undo.branch
   .hg/undo.desc
-  .hg/undo.dirstate
   .hg/undo.phaseroots
   .hg/wcache
   .hg/wcache/checkisexec (execbit !)
   .hg/wcache/checklink (symlink !)
   .hg/wcache/checklink-target (symlink !)
   .hg/wcache/manifestfulltextcache (reporevlogstore !)
   $ cd ..
 
 Non fncache repo:
 
   $ hg --config format.usefncache=False init bar
   $ cd bar
   $ mkdir tst.d
   $ echo foo > tst.d/Foo
   $ hg ci -Amfoo
   adding tst.d/Foo
   $ find .hg | sort
   .hg
   .hg/00changelog.i
   .hg/cache
   .hg/cache/branch2-served
   .hg/cache/rbc-names-v1
   .hg/cache/rbc-revs-v1
   .hg/dirstate
   .hg/fsmonitor.state (fsmonitor !)
   .hg/last-message.txt
   .hg/requires
   .hg/store
   .hg/store/00changelog.i
   .hg/store/00manifest.i
   .hg/store/data
   .hg/store/data/tst.d.hg
   .hg/store/data/tst.d.hg/_foo.i
   .hg/store/phaseroots
   .hg/store/requires
   .hg/store/undo
   .hg/store/undo.backupfiles
   .hg/store/undo.phaseroots
-  .hg/undo.backup.dirstate
   .hg/undo.bookmarks
   .hg/undo.branch
   .hg/undo.desc
-  .hg/undo.dirstate
   .hg/wcache
   .hg/wcache/checkisexec (execbit !)
   .hg/wcache/checklink (symlink !)
   .hg/wcache/checklink-target (symlink !)
   .hg/wcache/manifestfulltextcache (reporevlogstore !)
   $ cd ..
 
 Encoding of reserved / long paths in the store
 
   $ hg init r2
   $ cd r2
   $ cat <<EOF > .hg/hgrc
   > [ui]
   > portablefilenames = ignore
   > EOF
 
   $ hg import -q --bypass - <<EOF
   > # HG changeset patch
   > # User test
   > # Date 0 0
   > # Node ID 1c7a2f7cb77be1a0def34e4c7cabc562ad98fbd7
   > # Parent  0000000000000000000000000000000000000000
   > 1
   > 
   > diff --git a/12345678/12345678/12345678/12345678/12345678/12345678/12345678/12345/xxxxxxxxx-xxxxxxxxx-xxxxxxxxx-123456789-12.3456789-12345-ABCDEFGHIJKLMNOPRSTUVWXYZ-abcdefghjiklmnopqrstuvwxyz b/12345678/12345678/12345678/12345678/12345678/12345678/12345678/12345/xxxxxxxxx-xxxxxxxxx-xxxxxxxxx-123456789-12.3456789-12345-ABCDEFGHIJKLMNOPRSTUVWXYZ-abcdefghjiklmnopqrstuvwxyz
   > new file mode 100644
   > --- /dev/null
   > +++ b/12345678/12345678/12345678/12345678/12345678/12345678/12345678/12345/xxxxxxxxx-xxxxxxxxx-xxxxxxxxx-123456789-12.3456789-12345-ABCDEFGHIJKLMNOPRSTUVWXYZ-abcdefghjiklmnopqrstuvwxyz
   > @@ -0,0 +1,1 @@
   > +foo
   > diff --git a/AUX/SECOND/X.PRN/FOURTH/FI:FTH/SIXTH/SEVENTH/EIGHTH/NINETH/TENTH/ELEVENTH/LOREMIPSUM.TXT b/AUX/SECOND/X.PRN/FOURTH/FI:FTH/SIXTH/SEVENTH/EIGHTH/NINETH/TENTH/ELEVENTH/LOREMIPSUM.TXT
   > new file mode 100644
   > --- /dev/null
   > +++ b/AUX/SECOND/X.PRN/FOURTH/FI:FTH/SIXTH/SEVENTH/EIGHTH/NINETH/TENTH/ELEVENTH/LOREMIPSUM.TXT
   > @@ -0,0 +1,1 @@
   > +foo
   > diff --git a/Project Planning/Resources/AnotherLongDirectoryName/Followedbyanother/AndAnother/AndThenAnExtremelyLongFileName.txt b/Project Planning/Resources/AnotherLongDirectoryName/Followedbyanother/AndAnother/AndThenAnExtremelyLongFileName.txt
   > new file mode 100644
   > --- /dev/null
   > +++ b/Project Planning/Resources/AnotherLongDirectoryName/Followedbyanother/AndAnother/AndThenAnExtremelyLongFileName.txt	
   > @@ -0,0 +1,1 @@
   > +foo
   > diff --git a/bla.aux/prn/PRN/lpt/com3/nul/coma/foo.NUL/normal.c b/bla.aux/prn/PRN/lpt/com3/nul/coma/foo.NUL/normal.c
   > new file mode 100644
   > --- /dev/null
   > +++ b/bla.aux/prn/PRN/lpt/com3/nul/coma/foo.NUL/normal.c
   > @@ -0,0 +1,1 @@
   > +foo
   > diff --git a/enterprise/openesbaddons/contrib-imola/corba-bc/netbeansplugin/wsdlExtension/src/main/java/META-INF/services/org.netbeans.modules.xml.wsdl.bindingsupport.spi.ExtensibilityElementTemplateProvider b/enterprise/openesbaddons/contrib-imola/corba-bc/netbeansplugin/wsdlExtension/src/main/java/META-INF/services/org.netbeans.modules.xml.wsdl.bindingsupport.spi.ExtensibilityElementTemplateProvider
   > new file mode 100644
   > --- /dev/null
   > +++ b/enterprise/openesbaddons/contrib-imola/corba-bc/netbeansplugin/wsdlExtension/src/main/java/META-INF/services/org.netbeans.modules.xml.wsdl.bindingsupport.spi.ExtensibilityElementTemplateProvider
   > @@ -0,0 +1,1 @@
   > +foo
   > EOF
 
   $ find .hg/store -name *.i  | sort
   .hg/store/00changelog.i
   .hg/store/00manifest.i
   .hg/store/data/bla.aux/pr~6e/_p_r_n/lpt/co~6d3/nu~6c/coma/foo._n_u_l/normal.c.i
   .hg/store/dh/12345678/12345678/12345678/12345678/12345678/12345678/12345678/12345/xxxxxx168e07b38e65eff86ab579afaaa8e30bfbe0f35f.i
   .hg/store/dh/au~78/second/x.prn/fourth/fi~3afth/sixth/seventh/eighth/nineth/tenth/loremia20419e358ddff1bf8751e38288aff1d7c32ec05.i
   .hg/store/dh/enterpri/openesba/contrib-/corba-bc/netbeans/wsdlexte/src/main/java/org.net7018f27961fdf338a598a40c4683429e7ffb9743.i
   .hg/store/dh/project_/resource/anotherl/followed/andanoth/andthenanextremelylongfilename0d8e1f4187c650e2f1fdca9fd90f786bc0976b6b.i
 
   $ cd ..
 
 Aborting lock does not prevent fncache writes
 
   $ cat > exceptionext.py <<EOF
   > import os
   > from mercurial import commands, error, extensions
   > 
   > def lockexception(orig, vfs, lockname, wait, releasefn, *args, **kwargs):
   >     def releasewrap():
   >         l.held = False # ensure __del__ is a noop
   >         raise error.Abort(b"forced lock failure")
   >     l = orig(vfs, lockname, wait, releasewrap, *args, **kwargs)
   >     return l
   > 
   > def reposetup(ui, repo):
   >     extensions.wrapfunction(repo, '_lock', lockexception)
   > 
   > cmdtable = {}
   > 
   > # wrap "commit" command to prevent wlock from being '__del__()'-ed
   > # at the end of dispatching (for intentional "forced lcok failure")
   > def commitwrap(orig, ui, repo, *pats, **opts):
   >     repo = repo.unfiltered() # to use replaced repo._lock certainly
   >     wlock = repo.wlock()
   >     try:
   >         return orig(ui, repo, *pats, **opts)
   >     finally:
   >         # multiple 'relase()' is needed for complete releasing wlock,
   >         # because "forced" abort at last releasing store lock
   >         # prevents wlock from being released at same 'lockmod.release()'
   >         for i in range(wlock.held):
   >             wlock.release()
   > 
   > def extsetup(ui):
   >     extensions.wrapcommand(commands.table, b"commit", commitwrap)
   > EOF
   $ extpath=`pwd`/exceptionext.py
   $ hg init fncachetxn
   $ cd fncachetxn
   $ printf "[extensions]\nexceptionext=$extpath\n" >> .hg/hgrc
   $ touch y
   $ hg ci -qAm y
   abort: forced lock failure
   [255]
   $ cat .hg/store/fncache
   data/y.i
 
 Aborting transaction prevents fncache change
 
   $ cat > ../exceptionext.py <<EOF
   > import os
   > from mercurial import commands, error, extensions, localrepo
   > 
   > def wrapper(orig, self, *args, **kwargs):
   >     tr = orig(self, *args, **kwargs)
   >     def fail(tr):
   >         raise error.Abort(b"forced transaction failure")
   >     # zzz prefix to ensure it sorted after store.write
   >     tr.addfinalize(b'zzz-forcefails', fail)
   >     return tr
   > 
   > def uisetup(ui):
   >     extensions.wrapfunction(
   >         localrepo.localrepository, b'transaction', wrapper)
   > 
   > cmdtable = {}
   > 
   > EOF
 
 Clean cached version
   $ rm -f "${extpath}c"
   $ rm -Rf "`dirname $extpath`/__pycache__"
 
   $ touch z
   $ hg ci -qAm z
   transaction abort!
   rollback completed
   abort: forced transaction failure
   [255]
   $ cat .hg/store/fncache
   data/y.i
 
 Aborted transactions can be recovered later
 
   $ cat > ../exceptionext.py <<EOF
   > import os
   > import signal
   > from mercurial import (
   >   commands,
   >   error,
   >   extensions,
   >   localrepo,
   >   transaction,
   > )
   > 
   > def trwrapper(orig, self, *args, **kwargs):
   >     tr = orig(self, *args, **kwargs)
   >     def fail(tr):
   >         os.kill(os.getpid(), signal.SIGKILL)
   >     # zzz prefix to ensure it sorted after store.write
   >     tr.addfinalize(b'zzz-forcefails', fail)
   >     return tr
   > 
   > def uisetup(ui):
   >     extensions.wrapfunction(localrepo.localrepository, 'transaction',
   >                             trwrapper)
   > 
   > cmdtable = {}
   > 
   > EOF
 
 Clean cached versions
   $ rm -f "${extpath}c"
   $ rm -Rf "`dirname $extpath`/__pycache__"
 
   $ hg up -q 1
   $ touch z
 # Cannot rely on the return code value as chg use a different one.
 # So we use a `|| echo` trick
 # XXX-CHG fixing chg behavior would be nice here.
   $ hg ci -qAm z || echo "He's Dead, Jim." 2>/dev/null
   Killed (?)
   He's Dead, Jim.
   $ cat .hg/store/fncache | sort
   data/y.i
   data/z.i
   $ hg recover --verify
   rolling back interrupted transaction
   checking changesets
   checking manifests
   crosschecking files in changesets and manifests
   checking files
   checking dirstate
   checked 1 changesets with 1 changes to 1 files
   $ cat .hg/store/fncache
   data/y.i
 
   $ cd ..
 
 debugrebuildfncache does nothing unless repo has fncache requirement
 
   $ hg --config format.usefncache=false init nofncache
   $ cd nofncache
   $ hg debugrebuildfncache
   (not rebuilding fncache because repository does not support fncache)
 
   $ cd ..
 
 debugrebuildfncache works on empty repository
 
   $ hg init empty
   $ cd empty
   $ hg debugrebuildfncache
   fncache already up to date
   $ cd ..
 
 debugrebuildfncache on an up to date repository no-ops
 
   $ hg init repo
   $ cd repo
   $ echo initial > foo
   $ echo initial > .bar
   $ hg commit -A -m initial
   adding .bar
   adding foo
 
   $ cat .hg/store/fncache | sort
   data/.bar.i
   data/foo.i
 
   $ hg debugrebuildfncache
   fncache already up to date
 
 debugrebuildfncache restores deleted fncache file
 
   $ rm -f .hg/store/fncache
   $ hg debugrebuildfncache
   adding data/.bar.i
   adding data/foo.i
   2 items added, 0 removed from fncache
 
   $ cat .hg/store/fncache | sort
   data/.bar.i
   data/foo.i
 
 Rebuild after rebuild should no-op
 
   $ hg debugrebuildfncache
   fncache already up to date
 
 A single missing file should get restored, an extra file should be removed
 
   $ cat > .hg/store/fncache << EOF
   > data/foo.i
   > data/bad-entry.i
   > EOF
 
   $ hg debugrebuildfncache
   removing data/bad-entry.i
   adding data/.bar.i
   1 items added, 1 removed from fncache
 
   $ cat .hg/store/fncache | sort
   data/.bar.i
   data/foo.i
 
 debugrebuildfncache recovers from truncated line in fncache
 
   $ printf a > .hg/store/fncache
   $ hg debugrebuildfncache
   fncache does not ends with a newline
   adding data/.bar.i
   adding data/foo.i
   2 items added, 0 removed from fncache
 
   $ cat .hg/store/fncache | sort
   data/.bar.i
   data/foo.i
 
   $ cd ..
 
 Try a simple variation without dotencode to ensure fncache is ignorant of encoding
 
   $ hg --config format.dotencode=false init nodotencode
   $ cd nodotencode
   $ echo initial > foo
   $ echo initial > .bar
   $ hg commit -A -m initial
   adding .bar
   adding foo
 
   $ cat .hg/store/fncache | sort
   data/.bar.i
   data/foo.i
 
   $ rm .hg/store/fncache
   $ hg debugrebuildfncache
   adding data/.bar.i
   adding data/foo.i
   2 items added, 0 removed from fncache
 
   $ cat .hg/store/fncache | sort
   data/.bar.i
   data/foo.i
 
   $ cd ..
 
 In repositories that have accumulated a large number of files over time, the
 fncache file is going to be large. If we possibly can avoid loading it, so much the better.
 The cache should not loaded when committing changes to existing files, or when unbundling
 changesets that only contain changes to existing files:
 
   $ cat > fncacheloadwarn.py << EOF
   > from mercurial import extensions, localrepo
   > 
   > def extsetup(ui):
   >     def wrapstore(orig, requirements, *args):
   >         store = orig(requirements, *args)
   >         if b'store' in requirements and b'fncache' in requirements:
   >             instrumentfncachestore(store, ui)
   >         return store
   >     extensions.wrapfunction(localrepo, 'makestore', wrapstore)
   > 
   > def instrumentfncachestore(fncachestore, ui):
   >     class instrumentedfncache(type(fncachestore.fncache)):
   >         def _load(self):
   >             ui.warn(b'fncache load triggered!\n')
   >             super(instrumentedfncache, self)._load()
   >     fncachestore.fncache.__class__ = instrumentedfncache
   > EOF
 
   $ fncachextpath=`pwd`/fncacheloadwarn.py
   $ hg init nofncacheload
   $ cd nofncacheload
   $ printf "[extensions]\nfncacheloadwarn=$fncachextpath\n" >> .hg/hgrc
 
 A new file should trigger a load, as we'd want to update the fncache set in that case:
 
   $ touch foo
   $ hg ci -qAm foo
   fncache load triggered!
 
 But modifying that file should not:
 
   $ echo bar >> foo
   $ hg ci -qm foo
 
 If a transaction has been aborted, the zero-size truncated index file will
 not prevent the fncache from being loaded; rather than actually abort
 a transaction, we simulate the situation by creating a zero-size index file:
 
   $ touch .hg/store/data/bar.i
   $ touch bar
   $ hg ci -qAm bar
   fncache load triggered!
 
 Unbundling should follow the same rules; existing files should not cause a load:
 
 (loading during the clone is expected)
   $ hg clone -q . tobundle
   fncache load triggered!
   fncache load triggered!
 
   $ echo 'new line' > tobundle/bar
   $ hg -R tobundle ci -qm bar
   $ hg -R tobundle bundle -q barupdated.hg
   $ hg unbundle -q barupdated.hg
 
 but adding new files should:
 
   $ touch tobundle/newfile
   $ hg -R tobundle ci -qAm newfile
   $ hg -R tobundle bundle -q newfile.hg
   $ hg unbundle -q newfile.hg
   fncache load triggered!
 
   $ cd ..

tests/test-inherit-mode.t

@@ -1,184 +1,182 b''
 #require unix-permissions
 
 test that new files created in .hg inherit the permissions from .hg/store
 
   $ mkdir dir
 
 just in case somebody has a strange $TMPDIR
 
   $ chmod g-s dir
   $ cd dir
 
   $ cat >printmodes.py <<EOF
   > import os
   > import sys
   > 
   > allnames = []
   > isdir = {}
   > for root, dirs, files in os.walk(sys.argv[1]):
   >     for d in dirs:
   >         name = os.path.join(root, d)
   >         isdir[name] = 1
   >         allnames.append(name)
   >     for f in files:
   >         name = os.path.join(root, f)
   >         allnames.append(name)
   > allnames.sort()
   > for name in allnames:
   >     suffix = name in isdir and '/' or ''
   >     print('%05o %s%s' % (os.lstat(name).st_mode & 0o7777, name, suffix))
   > EOF
 
   $ cat >mode.py <<EOF
   > import os
   > import sys
   > print('%05o' % os.lstat(sys.argv[1]).st_mode)
   > EOF
 
   $ umask 077
 
   $ hg init repo
   $ cd repo
 
   $ chmod 0770 .hg/store .hg/cache .hg/wcache
 
 before commit
 store can be written by the group, other files cannot
 store is setgid
 
   $ "$PYTHON" ../printmodes.py .
   00700 ./.hg/
   00600 ./.hg/00changelog.i
   00770 ./.hg/cache/
   00600 ./.hg/requires
   00770 ./.hg/store/
   00600 ./.hg/store/requires
   00770 ./.hg/wcache/
 
   $ mkdir dir
   $ touch foo dir/bar
   $ hg ci -qAm 'add files'
 
 after commit
 working dir files can only be written by the owner
 files created in .hg can be written by the group
 (in particular, store/**, dirstate, branch cache file, undo files)
 new directories are setgid
 
   $ "$PYTHON" ../printmodes.py .
   00700 ./.hg/
   00600 ./.hg/00changelog.i
   00770 ./.hg/cache/
   00660 ./.hg/cache/branch2-served
   00660 ./.hg/cache/rbc-names-v1
   00660 ./.hg/cache/rbc-revs-v1
   00660 ./.hg/dirstate
   00660 ./.hg/fsmonitor.state (fsmonitor !)
   00660 ./.hg/last-message.txt
   00600 ./.hg/requires
   00770 ./.hg/store/
   00660 ./.hg/store/00changelog.i
   00660 ./.hg/store/00manifest.i
   00770 ./.hg/store/data/
   00770 ./.hg/store/data/dir/
   00660 ./.hg/store/data/dir/bar.i (reporevlogstore !)
   00660 ./.hg/store/data/foo.i (reporevlogstore !)
   00770 ./.hg/store/data/dir/bar/ (reposimplestore !)
   00660 ./.hg/store/data/dir/bar/b80de5d138758541c5f05265ad144ab9fa86d1db (reposimplestore !)
   00660 ./.hg/store/data/dir/bar/index (reposimplestore !)
   00770 ./.hg/store/data/foo/ (reposimplestore !)
   00660 ./.hg/store/data/foo/b80de5d138758541c5f05265ad144ab9fa86d1db (reposimplestore !)
   00660 ./.hg/store/data/foo/index (reposimplestore !)
   00660 ./.hg/store/fncache (repofncache !)
   00660 ./.hg/store/phaseroots
   00600 ./.hg/store/requires
   00660 ./.hg/store/undo
   00660 ./.hg/store/undo.backupfiles
   00660 ./.hg/store/undo.phaseroots
-  00660 ./.hg/undo.backup.dirstate
   00660 ./.hg/undo.bookmarks
   00660 ./.hg/undo.branch
   00660 ./.hg/undo.desc
-  00660 ./.hg/undo.dirstate
   00770 ./.hg/wcache/
   00711 ./.hg/wcache/checkisexec
   007.. ./.hg/wcache/checklink (re)
   00600 ./.hg/wcache/checklink-target
   00660 ./.hg/wcache/manifestfulltextcache (reporevlogstore !)
   00700 ./dir/
   00600 ./dir/bar
   00600 ./foo
 
   $ umask 007
   $ hg init ../push
 
 before push
 group can write everything
 
   $ "$PYTHON" ../printmodes.py ../push
   00770 ../push/.hg/
   00660 ../push/.hg/00changelog.i
   00770 ../push/.hg/cache/
   00660 ../push/.hg/requires
   00770 ../push/.hg/store/
   00660 ../push/.hg/store/requires
   00770 ../push/.hg/wcache/
 
   $ umask 077
   $ hg -q push ../push
 
 after push
 group can still write everything
 
   $ "$PYTHON" ../printmodes.py ../push
   00770 ../push/.hg/
   00660 ../push/.hg/00changelog.i
   00770 ../push/.hg/cache/
   00660 ../push/.hg/cache/branch2-base
   00660 ../push/.hg/cache/rbc-names-v1
   00660 ../push/.hg/cache/rbc-revs-v1
   00660 ../push/.hg/requires
   00770 ../push/.hg/store/
   00660 ../push/.hg/store/00changelog.i
   00660 ../push/.hg/store/00manifest.i
   00770 ../push/.hg/store/data/
   00770 ../push/.hg/store/data/dir/
   00660 ../push/.hg/store/data/dir/bar.i (reporevlogstore !)
   00660 ../push/.hg/store/data/foo.i (reporevlogstore !)
   00770 ../push/.hg/store/data/dir/bar/ (reposimplestore !)
   00660 ../push/.hg/store/data/dir/bar/b80de5d138758541c5f05265ad144ab9fa86d1db (reposimplestore !)
   00660 ../push/.hg/store/data/dir/bar/index (reposimplestore !)
   00770 ../push/.hg/store/data/foo/ (reposimplestore !)
   00660 ../push/.hg/store/data/foo/b80de5d138758541c5f05265ad144ab9fa86d1db (reposimplestore !)
   00660 ../push/.hg/store/data/foo/index (reposimplestore !)
   00660 ../push/.hg/store/fncache (repofncache !)
   00660 ../push/.hg/store/requires
   00660 ../push/.hg/store/undo
   00660 ../push/.hg/store/undo.backupfiles
   00660 ../push/.hg/store/undo.phaseroots
   00660 ../push/.hg/undo.bookmarks
   00660 ../push/.hg/undo.branch
   00660 ../push/.hg/undo.desc
   00770 ../push/.hg/wcache/
 
 
 Test that we don't lose the setgid bit when we call chmod.
 Not all systems support setgid directories (e.g. HFS+), so
 just check that directories have the same mode.
 
   $ cd ..
   $ hg init setgid
   $ cd setgid
   $ chmod g+rwx .hg/store
   $ chmod g+s .hg/store 2> /dev/null || true
   $ mkdir dir
   $ touch dir/file
   $ hg ci -qAm 'add dir/file'
   $ storemode=`"$PYTHON" ../mode.py .hg/store`
   $ dirmode=`"$PYTHON" ../mode.py .hg/store/data/dir`
   $ if [ "$storemode" != "$dirmode" ]; then
   >  echo "$storemode != $dirmode"
   > fi
   $ cd ..
 
   $ cd .. # g-s dir

tests/test-racy-mutations.t

@@ -1,125 +1,125 b''
 #testcases skip-detection fail-if-detected
 
 Test situations that "should" only be reproducible:
 - on networked filesystems, or
 - user using `hg debuglocks` to eliminate the lock file, or
 - something (that doesn't respect the lock file) writing to the .hg directory
 while we're running
 
 
 Initial setup
 -------------
 
   $ hg init base-repo
   $ cd base-repo
 
   $ cat > "$TESTTMP_FORWARD_SLASH/waitlock_editor.sh" <<EOF
   >     [ -n "\${WAITLOCK_ANNOUNCE:-}" ] && touch "\${WAITLOCK_ANNOUNCE}"
   >     f="\${WAITLOCK_FILE}"
   >     start=\`date +%s\`
   >     timeout=5
   >     "$RUNTESTDIR_FORWARD_SLASH/testlib/wait-on-file" "\$timeout" "\$f"
   >     if [ \$# -gt 1 ]; then
   >         cat "\$@"
   >     fi
   > EOF
 
 Things behave differently if we don't already have a 00changelog.i file when
 this all starts, so let's make one.
 
   $ echo r0 > r0
   $ hg commit -qAm 'r0'
 
   $ cd ..
   $ cp -R base-repo main-client
   $ cp -R base-repo racing-client
 
   $ mkdir sync
   $ EDITOR_STARTED="$TESTTMP_FORWARD_SLASH/sync/.editor_started"
   $ MISCHIEF_MANAGED="$TESTTMP_FORWARD_SLASH/sync/.mischief_managed"
   $ JOBS_FINISHED="$TESTTMP_FORWARD_SLASH/sync/.jobs_finished"
 
 Actual test
 -----------
 
 Start an hg commit that will take a while
 
   $ cd main-client
 
 #if fail-if-detected
   $ cat >> $HGRCPATH << EOF
   > [debug]
   > revlog.verifyposition.changelog = fail
   > EOF
 #endif
 
   $ echo foo > foo
   $ (
   >    unset HGEDITOR;
   >    WAITLOCK_ANNOUNCE="${EDITOR_STARTED}" \
   >    WAITLOCK_FILE="${MISCHIEF_MANAGED}" \
   >    hg commit -qAm 'r1 (foo)' --edit foo \
   >    --config ui.editor="sh $TESTTMP_FORWARD_SLASH/waitlock_editor.sh" \
   >    > .foo_commit_out 2>&1 ;\
   >    touch "${JOBS_FINISHED}"
   > ) &
 
 Wait for the "editor" to actually start
   $ sh "$RUNTESTDIR_FORWARD_SLASH/testlib/wait-on-file" 5 "${EDITOR_STARTED}"
 
 
 Do a concurrent edition
   $ cd ../racing-client
   $ touch ../pre-race
   $ sleep 1
   $ echo bar > bar
   $ hg --repository ../racing-client commit -qAm 'r2 (bar)' bar
   $ hg --repository ../racing-client debugrevlogindex -c
      rev linkrev nodeid       p1           p2
        0       0 222799e2f90b 000000000000 000000000000
        1       1 6f124f6007a0 222799e2f90b 000000000000
 
 We simulate an network FS race by overwriting raced repo content with the new
 content of the files changed in the racing repository
 
   $ for x in `find . -type f -newer ../pre-race`; do
   >    cp $x ../main-client/$x
   > done
   $ cd ../main-client
 
 Awaken the editor from that first commit
   $ touch "${MISCHIEF_MANAGED}"
 And wait for it to finish
   $ WAITLOCK_FILE="${JOBS_FINISHED}" sh "$TESTTMP_FORWARD_SLASH/waitlock_editor.sh"
 
 #if skip-detection
 (Ensure there was no output)
   $ cat .foo_commit_out
 And observe a corrupted repository -- rev 2's linkrev is 1, which should never
 happen for the changelog (the linkrev should always refer to itself).
   $ hg debugrevlogindex -c
      rev linkrev nodeid       p1           p2
        0       0 222799e2f90b 000000000000 000000000000
        1       1 6f124f6007a0 222799e2f90b 000000000000
        2       1 ac80e6205bb2 222799e2f90b 000000000000
 #endif
 
 #if fail-if-detected
   $ cat .foo_commit_out
+  note: commit message saved in .hg/last-message.txt
+  note: use 'hg commit --logfile .hg/last-message.txt --edit' to reuse it
   transaction abort!
   rollback completed
-  note: commit message saved in .hg/last-message.txt
-  note: use 'hg commit --logfile .hg/last-message.txt --edit' to reuse it
   abort: 00changelog.i: file cursor at position 249, expected 121
 And no corruption in the changelog.
   $ hg debugrevlogindex -c
      rev linkrev nodeid       p1           p2
        0       0 222799e2f90b 000000000000 000000000000
        1       1 6f124f6007a0 222799e2f90b 000000000000 (missing-correct-output !)
 And, because of transactions, there's none in the manifestlog either.
   $ hg debugrevlogindex -m
      rev linkrev nodeid       p1           p2
        0       0 7b7020262a56 000000000000 000000000000
        1       1 ad3fe36d86d9 7b7020262a56 000000000000
 #endif