largefiles: track if a matcher was tampered with

This is used to make sure rust fast path is not taken for the
modified matchers.

hgext/largefiles/overrides.py

@@ -1,1924 +1,1932 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)
+    m._was_tampered_with = True
     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)
+    m._was_tampered_with = True
     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, '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, '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, '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, '_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(dirstate.dirstate, 'running_status')
 @contextlib.contextmanager
 def running_status(orig, self, repo):
     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):
             if sub_dirstate is None:
                 yield
             else:
                 with sub_dirstate.running_status(repo):
                     yield
     finally:
         self._sub_dirstate = pre
 
 
 @eh.wrapfunction(subrepo.hgsubrepo, '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, '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)
 
+        m._was_tampered_with = True
+
         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, 'matchandpats', overridematchandpats
     )
     wrappedmakefilematcher = extensions.wrappedfunction(
         logcmdutil, '_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, '_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, '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, 'recordupdates')
 def mergerecordupdates(orig, repo, actions, branchmerge, getfiledata):
     if MERGE_ACTION_LARGEFILE_MARK_REMOVED in actions:
         lfdirstate = lfutil.openlfdirstate(repo.ui, 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)
 
     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, '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, '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, '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, '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)
+            m._was_tampered_with = True
             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, 'copyfile', overridecopyfile):
             with extensions.wrappedfunction(scmutil, '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, '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(), repo.dirstate.running_status(repo):
         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)
+            m._was_tampered_with = True
 
             # 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, '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, '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, '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 hasattr(repo, '_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, '_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, 'archive')
 def hgwebarchive(orig, web):
     with lfstatus(web.repo):
         return orig(web)
 
 
 @eh.wrapfunction(archival, '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 (
                 hasattr(sub, '_repo')
                 and lfstatus(sub._repo)
                 or util.nullcontextmanager()
             ):
                 sub.archive(archiver, subprefix, submatch)
 
     archiver.done()
 
 
 @eh.wrapfunction(subrepo.hgsubrepo, 'archive')
 def hgsubrepoarchive(orig, repo, archiver, prefix, match=None, decode=True):
     lfenabled = hasattr(repo._repo, '_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 (
             hasattr(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, '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, 'postcommitstatus')
 def postcommitstatus(orig, repo, *args, **kwargs):
     with lfstatus(repo):
         return orig(repo, *args, **kwargs)
 
 
 @eh.wrapfunction(cmdutil, '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, 'addremove')
 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, open_tr=open_tr)
 
     # open the transaction and changing_files context
     if open_tr is not None:
         open_tr()
 
     # Get the list of missing largefiles so we can remove them
     with repo.dirstate.running_status(repo):
         lfdirstate = lfutil.openlfdirstate(repo.ui, repo)
         unsure, s, mtime_boundary = lfdirstate.status(
             matchmod.always(),
             subrepos=[],
             ignored=False,
             clean=False,
             unknown=False,
         )
 
     # 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)
+        m._was_tampered_with = True
 
         # 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, 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):
     ctx = logcmdutil.revsingle(repo, opts.get('rev'))
     err = 1
     notbad = set()
     m = scmutil.match(ctx, (file1,) + pats, pycompat.byteskwargs(opts))
+    m._was_tampered_with = True
     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('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('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, '_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(), repo.dirstate.changing_parents(repo):
         # 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")
         with repo.dirstate.running_status(repo):
             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:
             entry = lfdirstate.get_entry(lfile)
             lfdirstate.hacky_extension_update_file(
                 lfile,
                 wc_tracked=entry.tracked,
                 p1_tracked=entry.p1_tracked,
                 p2_info=entry.p2_info,
                 possibly_dirty=True,
             )
         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'
             )
         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)
 
         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, '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, 'preservedrequirements')
 @eh.wrapfunction(upgrade_actions, '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, '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/match.py

@@ -1,1706 +1,1712 b''
 # match.py - filename matching
 #
 #  Copyright 2008, 2009 Olivia Mackall <[e-mail]> and others
 #
 # This software may be used and distributed according to the terms of the
 # GNU General Public License version 2 or any later version.
 
 
 import bisect
 import copy
 import itertools
 import os
 import re
 
 from .i18n import _
 from .pycompat import open
 from . import (
     encoding,
     error,
     pathutil,
     policy,
     pycompat,
     util,
 )
 from .utils import stringutil
 
 rustmod = policy.importrust('dirstate')
 
 allpatternkinds = (
     b're',
     b'glob',
     b'path',
     b'filepath',
     b'relglob',
     b'relpath',
     b'relre',
     b'rootglob',
     b'listfile',
     b'listfile0',
     b'set',
     b'include',
     b'subinclude',
     b'rootfilesin',
 )
 cwdrelativepatternkinds = (b'relpath', b'glob')
 
 propertycache = util.propertycache
 
 
 def _rematcher(regex):
     """compile the regexp with the best available regexp engine and return a
     matcher function"""
     m = util.re.compile(regex)
     try:
         # slightly faster, provided by facebook's re2 bindings
         return m.test_match
     except AttributeError:
         return m.match
 
 
 def _expandsets(cwd, kindpats, ctx=None, listsubrepos=False, badfn=None):
     '''Returns the kindpats list with the 'set' patterns expanded to matchers'''
     matchers = []
     other = []
 
     for kind, pat, source in kindpats:
         if kind == b'set':
             if ctx is None:
                 raise error.ProgrammingError(
                     b"fileset expression with no context"
                 )
             matchers.append(ctx.matchfileset(cwd, pat, badfn=badfn))
 
             if listsubrepos:
                 for subpath in ctx.substate:
                     sm = ctx.sub(subpath).matchfileset(cwd, pat, badfn=badfn)
                     pm = prefixdirmatcher(subpath, sm, badfn=badfn)
                     matchers.append(pm)
 
             continue
         other.append((kind, pat, source))
     return matchers, other
 
 
 def _expandsubinclude(kindpats, root):
     """Returns the list of subinclude matcher args and the kindpats without the
     subincludes in it."""
     relmatchers = []
     other = []
 
     for kind, pat, source in kindpats:
         if kind == b'subinclude':
             sourceroot = pathutil.dirname(util.normpath(source))
             pat = util.pconvert(pat)
             path = pathutil.join(sourceroot, pat)
 
             newroot = pathutil.dirname(path)
             matcherargs = (newroot, b'', [], [b'include:%s' % path])
 
             prefix = pathutil.canonpath(root, root, newroot)
             if prefix:
                 prefix += b'/'
             relmatchers.append((prefix, matcherargs))
         else:
             other.append((kind, pat, source))
 
     return relmatchers, other
 
 
 def _kindpatsalwaysmatch(kindpats):
     """Checks whether the kindspats match everything, as e.g.
     'relpath:.' does.
     """
     for kind, pat, source in kindpats:
         if pat != b'' or kind not in [b'relpath', b'glob']:
             return False
     return True
 
 
 def _buildkindpatsmatcher(
     matchercls,
     root,
     cwd,
     kindpats,
     ctx=None,
     listsubrepos=False,
     badfn=None,
 ):
     matchers = []
     fms, kindpats = _expandsets(
         cwd,
         kindpats,
         ctx=ctx,
         listsubrepos=listsubrepos,
         badfn=badfn,
     )
     if kindpats:
         m = matchercls(root, kindpats, badfn=badfn)
         matchers.append(m)
     if fms:
         matchers.extend(fms)
     if not matchers:
         return nevermatcher(badfn=badfn)
     if len(matchers) == 1:
         return matchers[0]
     return unionmatcher(matchers)
 
 
 def match(
     root,
     cwd,
     patterns=None,
     include=None,
     exclude=None,
     default=b'glob',
     auditor=None,
     ctx=None,
     listsubrepos=False,
     warn=None,
     badfn=None,
     icasefs=False,
 ):
     r"""build an object to match a set of file patterns
 
     arguments:
     root - the canonical root of the tree you're matching against
     cwd - the current working directory, if relevant
     patterns - patterns to find
     include - patterns to include (unless they are excluded)
     exclude - patterns to exclude (even if they are included)
     default - if a pattern in patterns has no explicit type, assume this one
     auditor - optional path auditor
     ctx - optional changecontext
     listsubrepos - if True, recurse into subrepositories
     warn - optional function used for printing warnings
     badfn - optional bad() callback for this matcher instead of the default
     icasefs - make a matcher for wdir on case insensitive filesystems, which
         normalizes the given patterns to the case in the filesystem
 
     a pattern is one of:
     'glob:<glob>' - a glob relative to cwd
     're:<regexp>' - a regular expression
     'path:<path>' - a path relative to repository root, which is matched
                     recursively
     'filepath:<path>' - an exact path to a single file, relative to the
                         repository root
     'rootfilesin:<path>' - a path relative to repository root, which is
                     matched non-recursively (will not match subdirectories)
     'relglob:<glob>' - an unrooted glob (*.c matches C files in all dirs)
     'relpath:<path>' - a path relative to cwd
     'relre:<regexp>' - a regexp that needn't match the start of a name
     'set:<fileset>' - a fileset expression
     'include:<path>' - a file of patterns to read and include
     'subinclude:<path>' - a file of patterns to match against files under
                           the same directory
     '<something>' - a pattern of the specified default type
 
     >>> def _match(root, *args, **kwargs):
     ...     return match(util.localpath(root), *args, **kwargs)
 
     Usually a patternmatcher is returned:
     >>> _match(b'/foo', b'.', [br're:.*\.c$', b'path:foo/a', b'*.py'])
     <patternmatcher patterns='[^/]*\\.py$|foo/a(?:/|$)|.*\\.c$'>
 
     Combining 'patterns' with 'include' (resp. 'exclude') gives an
     intersectionmatcher (resp. a differencematcher):
     >>> type(_match(b'/foo', b'.', [br're:.*\.c$'], include=[b'path:lib']))
     <class 'mercurial.match.intersectionmatcher'>
     >>> type(_match(b'/foo', b'.', [br're:.*\.c$'], exclude=[b'path:build']))
     <class 'mercurial.match.differencematcher'>
 
     Notice that, if 'patterns' is empty, an alwaysmatcher is returned:
     >>> _match(b'/foo', b'.', [])
     <alwaysmatcher>
 
     The 'default' argument determines which kind of pattern is assumed if a
     pattern has no prefix:
     >>> _match(b'/foo', b'.', [br'.*\.c$'], default=b're')
     <patternmatcher patterns='.*\\.c$'>
     >>> _match(b'/foo', b'.', [b'main.py'], default=b'relpath')
     <patternmatcher patterns='main\\.py(?:/|$)'>
     >>> _match(b'/foo', b'.', [b'main.py'], default=b're')
     <patternmatcher patterns='main.py'>
 
     The primary use of matchers is to check whether a value (usually a file
     name) matches againset one of the patterns given at initialization. There
     are two ways of doing this check.
 
     >>> m = _match(b'/foo', b'', [br're:.*\.c$', b'relpath:a'])
 
     1. Calling the matcher with a file name returns True if any pattern
     matches that file name:
     >>> m(b'a')
     True
     >>> m(b'main.c')
     True
     >>> m(b'test.py')
     False
 
     2. Using the exact() method only returns True if the file name matches one
     of the exact patterns (i.e. not re: or glob: patterns):
     >>> m.exact(b'a')
     True
     >>> m.exact(b'main.c')
     False
     """
     assert os.path.isabs(root)
     cwd = os.path.join(root, util.localpath(cwd))
     normalize = _donormalize
     if icasefs:
         dirstate = ctx.repo().dirstate
         dsnormalize = dirstate.normalize
 
         def normalize(patterns, default, root, cwd, auditor, warn):
             kp = _donormalize(patterns, default, root, cwd, auditor, warn)
             kindpats = []
             for kind, pats, source in kp:
                 if kind not in (b're', b'relre'):  # regex can't be normalized
                     p = pats
                     pats = dsnormalize(pats)
 
                     # Preserve the original to handle a case only rename.
                     if p != pats and p in dirstate:
                         kindpats.append((kind, p, source))
 
                 kindpats.append((kind, pats, source))
             return kindpats
 
     if patterns:
         kindpats = normalize(patterns, default, root, cwd, auditor, warn)
         if _kindpatsalwaysmatch(kindpats):
             m = alwaysmatcher(badfn)
         else:
             m = _buildkindpatsmatcher(
                 patternmatcher,
                 root,
                 cwd,
                 kindpats,
                 ctx=ctx,
                 listsubrepos=listsubrepos,
                 badfn=badfn,
             )
     else:
         # It's a little strange that no patterns means to match everything.
         # Consider changing this to match nothing (probably using nevermatcher).
         m = alwaysmatcher(badfn)
 
     if include:
         kindpats = normalize(include, b'glob', root, cwd, auditor, warn)
         im = _buildkindpatsmatcher(
             includematcher,
             root,
             cwd,
             kindpats,
             ctx=ctx,
             listsubrepos=listsubrepos,
             badfn=None,
         )
         m = intersectmatchers(m, im)
     if exclude:
         kindpats = normalize(exclude, b'glob', root, cwd, auditor, warn)
         em = _buildkindpatsmatcher(
             includematcher,
             root,
             cwd,
             kindpats,
             ctx=ctx,
             listsubrepos=listsubrepos,
             badfn=None,
         )
         m = differencematcher(m, em)
     return m
 
 
 def exact(files, badfn=None):
     return exactmatcher(files, badfn=badfn)
 
 
 def always(badfn=None):
     return alwaysmatcher(badfn)
 
 
 def never(badfn=None):
     return nevermatcher(badfn)
 
 
 def badmatch(match, badfn):
     """Make a copy of the given matcher, replacing its bad method with the given
     one.
     """
     m = copy.copy(match)
     m.bad = badfn
     return m
 
 
 def _donormalize(patterns, default, root, cwd, auditor=None, warn=None):
     """Convert 'kind:pat' from the patterns list to tuples with kind and
     normalized and rooted patterns and with listfiles expanded."""
     kindpats = []
     kinds_to_normalize = (
         b'relglob',
         b'path',
         b'filepath',
         b'rootfilesin',
         b'rootglob',
     )
 
     for kind, pat in [_patsplit(p, default) for p in patterns]:
         if kind in cwdrelativepatternkinds:
             pat = pathutil.canonpath(root, cwd, pat, auditor=auditor)
         elif kind in kinds_to_normalize:
             pat = util.normpath(pat)
         elif kind in (b'listfile', b'listfile0'):
             try:
                 files = util.readfile(pat)
                 if kind == b'listfile0':
                     files = files.split(b'\0')
                 else:
                     files = files.splitlines()
                 files = [f for f in files if f]
             except EnvironmentError:
                 raise error.Abort(_(b"unable to read file list (%s)") % pat)
             for k, p, source in _donormalize(
                 files, default, root, cwd, auditor, warn
             ):
                 kindpats.append((k, p, pat))
             continue
         elif kind == b'include':
             try:
                 fullpath = os.path.join(root, util.localpath(pat))
                 includepats = readpatternfile(fullpath, warn)
                 for k, p, source in _donormalize(
                     includepats, default, root, cwd, auditor, warn
                 ):
                     kindpats.append((k, p, source or pat))
             except error.Abort as inst:
                 raise error.Abort(
                     b'%s: %s'
                     % (
                         pat,
                         inst.message,
                     )
                 )
             except IOError as inst:
                 if warn:
                     warn(
                         _(b"skipping unreadable pattern file '%s': %s\n")
                         % (pat, stringutil.forcebytestr(inst.strerror))
                     )
             continue
         # else: re or relre - which cannot be normalized
         kindpats.append((kind, pat, b''))
     return kindpats
 
 
 class basematcher:
     def __init__(self, badfn=None):
+        self._was_tampered_with = False
         if badfn is not None:
             self.bad = badfn
 
+    def was_tampered_with(self):
+        # [_was_tampered_with] is used to track if when extensions changed the matcher
+        # behavior (crazy stuff!), so we disable the rust fast path.
+        return self._was_tampered_with
+
     def __call__(self, fn):
         return self.matchfn(fn)
 
     # Callbacks related to how the matcher is used by dirstate.walk.
     # Subscribers to these events must monkeypatch the matcher object.
     def bad(self, f, msg):
         """Callback from dirstate.walk for each explicit file that can't be
         found/accessed, with an error message."""
 
     # If an traversedir is set, it will be called when a directory discovered
     # by recursive traversal is visited.
     traversedir = None
 
     @propertycache
     def _files(self):
         return []
 
     def files(self):
         """Explicitly listed files or patterns or roots:
         if no patterns or .always(): empty list,
         if exact: list exact files,
         if not .anypats(): list all files and dirs,
         else: optimal roots"""
         return self._files
 
     @propertycache
     def _fileset(self):
         return set(self._files)
 
     def exact(self, f):
         '''Returns True if f is in .files().'''
         return f in self._fileset
 
     def matchfn(self, f):
         return False
 
     def visitdir(self, dir):
         """Decides whether a directory should be visited based on whether it
         has potential matches in it or one of its subdirectories. This is
         based on the match's primary, included, and excluded patterns.
 
         Returns the string 'all' if the given directory and all subdirectories
         should be visited. Otherwise returns True or False indicating whether
         the given directory should be visited.
         """
         return True
 
     def visitchildrenset(self, dir):
         """Decides whether a directory should be visited based on whether it
         has potential matches in it or one of its subdirectories, and
         potentially lists which subdirectories of that directory should be
         visited. This is based on the match's primary, included, and excluded
         patterns.
 
         This function is very similar to 'visitdir', and the following mapping
         can be applied:
 
              visitdir | visitchildrenlist
             ----------+-------------------
              False    | set()
              'all'    | 'all'
              True     | 'this' OR non-empty set of subdirs -or files- to visit
 
         Example:
           Assume matchers ['path:foo/bar', 'rootfilesin:qux'], we would return
           the following values (assuming the implementation of visitchildrenset
           is capable of recognizing this; some implementations are not).
 
           '' -> {'foo', 'qux'}
           'baz' -> set()
           'foo' -> {'bar'}
           # Ideally this would be 'all', but since the prefix nature of matchers
           # is applied to the entire matcher, we have to downgrade this to
           # 'this' due to the non-prefix 'rootfilesin'-kind matcher being mixed
           # in.
           'foo/bar' -> 'this'
           'qux' -> 'this'
 
         Important:
           Most matchers do not know if they're representing files or
           directories. They see ['path:dir/f'] and don't know whether 'f' is a
           file or a directory, so visitchildrenset('dir') for most matchers will
           return {'f'}, but if the matcher knows it's a file (like exactmatcher
           does), it may return 'this'. Do not rely on the return being a set
           indicating that there are no files in this dir to investigate (or
           equivalently that if there are files to investigate in 'dir' that it
           will always return 'this').
         """
         return b'this'
 
     def always(self):
         """Matcher will match everything and .files() will be empty --
         optimization might be possible."""
         return False
 
     def isexact(self):
         """Matcher will match exactly the list of files in .files() --
         optimization might be possible."""
         return False
 
     def prefix(self):
         """Matcher will match the paths in .files() recursively --
         optimization might be possible."""
         return False
 
     def anypats(self):
         """None of .always(), .isexact(), and .prefix() is true --
         optimizations will be difficult."""
         return not self.always() and not self.isexact() and not self.prefix()
 
 
 class alwaysmatcher(basematcher):
     '''Matches everything.'''
 
     def __init__(self, badfn=None):
         super(alwaysmatcher, self).__init__(badfn)
 
     def always(self):
         return True
 
     def matchfn(self, f):
         return True
 
     def visitdir(self, dir):
         return b'all'
 
     def visitchildrenset(self, dir):
         return b'all'
 
     def __repr__(self):
         return r'<alwaysmatcher>'
 
 
 class nevermatcher(basematcher):
     '''Matches nothing.'''
 
     def __init__(self, badfn=None):
         super(nevermatcher, self).__init__(badfn)
 
     # It's a little weird to say that the nevermatcher is an exact matcher
     # or a prefix matcher, but it seems to make sense to let callers take
     # fast paths based on either. There will be no exact matches, nor any
     # prefixes (files() returns []), so fast paths iterating over them should
     # be efficient (and correct).
     def isexact(self):
         return True
 
     def prefix(self):
         return True
 
     def visitdir(self, dir):
         return False
 
     def visitchildrenset(self, dir):
         return set()
 
     def __repr__(self):
         return r'<nevermatcher>'
 
 
 class predicatematcher(basematcher):
     """A matcher adapter for a simple boolean function"""
 
     def __init__(self, predfn, predrepr=None, badfn=None):
         super(predicatematcher, self).__init__(badfn)
         self.matchfn = predfn
         self._predrepr = predrepr
 
     @encoding.strmethod
     def __repr__(self):
         s = stringutil.buildrepr(self._predrepr) or pycompat.byterepr(
             self.matchfn
         )
         return b'<predicatenmatcher pred=%s>' % s
 
 
 def path_or_parents_in_set(path, prefix_set):
     """Returns True if `path` (or any parent of `path`) is in `prefix_set`."""
     l = len(prefix_set)
     if l == 0:
         return False
     if path in prefix_set:
         return True
     # If there's more than 5 paths in prefix_set, it's *probably* quicker to
     # "walk up" the directory hierarchy instead, with the assumption that most
     # directory hierarchies are relatively shallow and hash lookup is cheap.
     if l > 5:
         return any(
             parentdir in prefix_set for parentdir in pathutil.finddirs(path)
         )
 
     # FIXME: Ideally we'd never get to this point if this is the case - we'd
     # recognize ourselves as an 'always' matcher and skip this.
     if b'' in prefix_set:
         return True
 
     sl = ord(b'/')
 
     # We already checked that path isn't in prefix_set exactly, so
     # `path[len(pf)] should never raise IndexError.
     return any(path.startswith(pf) and path[len(pf)] == sl for pf in prefix_set)
 
 
 class patternmatcher(basematcher):
     r"""Matches a set of (kind, pat, source) against a 'root' directory.
 
     >>> kindpats = [
     ...     (b're', br'.*\.c$', b''),
     ...     (b'path', b'foo/a', b''),
     ...     (b'relpath', b'b', b''),
     ...     (b'glob', b'*.h', b''),
     ... ]
     >>> m = patternmatcher(b'foo', kindpats)
     >>> m(b'main.c')  # matches re:.*\.c$
     True
     >>> m(b'b.txt')
     False
     >>> m(b'foo/a')  # matches path:foo/a
     True
     >>> m(b'a')  # does not match path:b, since 'root' is 'foo'
     False
     >>> m(b'b')  # matches relpath:b, since 'root' is 'foo'
     True
     >>> m(b'lib.h')  # matches glob:*.h
     True
 
     >>> m.files()
     [b'', b'foo/a', b'', b'b']
     >>> m.exact(b'foo/a')
     True
     >>> m.exact(b'b')
     True
     >>> m.exact(b'lib.h')  # exact matches are for (rel)path kinds
     False
     """
 
     def __init__(self, root, kindpats, badfn=None):
         super(patternmatcher, self).__init__(badfn)
         kindpats.sort()
 
         roots, dirs, parents = _rootsdirsandparents(kindpats)
         self._files = _explicitfiles(kindpats)
         self._dirs_explicit = set(dirs)
         self._dirs = parents
         self._prefix = _prefix(kindpats)
         self._pats, self._matchfn = _buildmatch(kindpats, b'$', root)
 
     def matchfn(self, fn):
         if fn in self._fileset:
             return True
         return self._matchfn(fn)
 
     def visitdir(self, dir):
         if self._prefix and dir in self._fileset:
             return b'all'
         return (
             dir in self._dirs
             or path_or_parents_in_set(dir, self._fileset)
             or path_or_parents_in_set(dir, self._dirs_explicit)
         )
 
     def visitchildrenset(self, dir):
         ret = self.visitdir(dir)
         if ret is True:
             return b'this'
         elif not ret:
             return set()
         assert ret == b'all'
         return b'all'
 
     def prefix(self):
         return self._prefix
 
     @encoding.strmethod
     def __repr__(self):
         return b'<patternmatcher patterns=%r>' % pycompat.bytestr(self._pats)
 
 
 # This is basically a reimplementation of pathutil.dirs that stores the
 # children instead of just a count of them, plus a small optional optimization
 # to avoid some directories we don't need.
 class _dirchildren:
     def __init__(self, paths, onlyinclude=None):
         self._dirs = {}
         self._onlyinclude = onlyinclude or []
         addpath = self.addpath
         for f in paths:
             addpath(f)
 
     def addpath(self, path):
         if path == b'':
             return
         dirs = self._dirs
         findsplitdirs = _dirchildren._findsplitdirs
         for d, b in findsplitdirs(path):
             if d not in self._onlyinclude:
                 continue
             dirs.setdefault(d, set()).add(b)
 
     @staticmethod
     def _findsplitdirs(path):
         # yields (dirname, basename) tuples, walking back to the root.  This is
         # very similar to pathutil.finddirs, except:
         #  - produces a (dirname, basename) tuple, not just 'dirname'
         # Unlike manifest._splittopdir, this does not suffix `dirname` with a
         # slash.
         oldpos = len(path)
         pos = path.rfind(b'/')
         while pos != -1:
             yield path[:pos], path[pos + 1 : oldpos]
             oldpos = pos
             pos = path.rfind(b'/', 0, pos)
         yield b'', path[:oldpos]
 
     def get(self, path):
         return self._dirs.get(path, set())
 
 
 class includematcher(basematcher):
     def __init__(self, root, kindpats, badfn=None):
         super(includematcher, self).__init__(badfn)
         if rustmod is not None:
             # We need to pass the patterns to Rust because they can contain
             # patterns from the user interface
             self._kindpats = kindpats
         self._pats, self.matchfn = _buildmatch(kindpats, b'(?:/|$)', root)
         self._prefix = _prefix(kindpats)
         roots, dirs, parents = _rootsdirsandparents(kindpats)
         # roots are directories which are recursively included.
         self._roots = set(roots)
         # dirs are directories which are non-recursively included.
         self._dirs = set(dirs)
         # parents are directories which are non-recursively included because
         # they are needed to get to items in _dirs or _roots.
         self._parents = parents
 
     def visitdir(self, dir):
         if self._prefix and dir in self._roots:
             return b'all'
         return (
             dir in self._dirs
             or dir in self._parents
             or path_or_parents_in_set(dir, self._roots)
         )
 
     @propertycache
     def _allparentschildren(self):
         # It may seem odd that we add dirs, roots, and parents, and then
         # restrict to only parents. This is to catch the case of:
         #   dirs = ['foo/bar']
         #   parents = ['foo']
         # if we asked for the children of 'foo', but had only added
         # self._parents, we wouldn't be able to respond ['bar'].
         return _dirchildren(
             itertools.chain(self._dirs, self._roots, self._parents),
             onlyinclude=self._parents,
         )
 
     def visitchildrenset(self, dir):
         if self._prefix and dir in self._roots:
             return b'all'
         # Note: this does *not* include the 'dir in self._parents' case from
         # visitdir, that's handled below.
         if (
             b'' in self._roots
             or dir in self._dirs
             or path_or_parents_in_set(dir, self._roots)
         ):
             return b'this'
 
         if dir in self._parents:
             return self._allparentschildren.get(dir) or set()
         return set()
 
     @encoding.strmethod
     def __repr__(self):
         return b'<includematcher includes=%r>' % pycompat.bytestr(self._pats)
 
 
 class exactmatcher(basematcher):
     r"""Matches the input files exactly. They are interpreted as paths, not
     patterns (so no kind-prefixes).
 
     >>> m = exactmatcher([b'a.txt', br're:.*\.c$'])
     >>> m(b'a.txt')
     True
     >>> m(b'b.txt')
     False
 
     Input files that would be matched are exactly those returned by .files()
     >>> m.files()
     ['a.txt', 're:.*\\.c$']
 
     So pattern 're:.*\.c$' is not considered as a regex, but as a file name
     >>> m(b'main.c')
     False
     >>> m(br're:.*\.c$')
     True
     """
 
     def __init__(self, files, badfn=None):
         super(exactmatcher, self).__init__(badfn)
 
         if isinstance(files, list):
             self._files = files
         else:
             self._files = list(files)
 
     matchfn = basematcher.exact
 
     @propertycache
     def _dirs(self):
         return set(pathutil.dirs(self._fileset))
 
     def visitdir(self, dir):
         return dir in self._dirs
 
     @propertycache
     def _visitchildrenset_candidates(self):
         """A memoized set of candidates for visitchildrenset."""
         return self._fileset | self._dirs - {b''}
 
     @propertycache
     def _sorted_visitchildrenset_candidates(self):
         """A memoized sorted list of candidates for visitchildrenset."""
         return sorted(self._visitchildrenset_candidates)
 
     def visitchildrenset(self, dir):
         if not self._fileset or dir not in self._dirs:
             return set()
 
         if dir == b'':
             candidates = self._visitchildrenset_candidates
         else:
             candidates = self._sorted_visitchildrenset_candidates
             d = dir + b'/'
             # Use bisect to find the first element potentially starting with d
             # (i.e. >= d). This should always find at least one element (we'll
             # assert later if this is not the case).
             first = bisect.bisect_left(candidates, d)
             # We need a representation of the first element that is > d that
             # does not start with d, so since we added a `/` on the end of dir,
             # we'll add whatever comes after slash (we could probably assume
             # that `0` is after `/`, but let's not) to the end of dir instead.
             dnext = dir + encoding.strtolocal(chr(ord(b'/') + 1))
             # Use bisect to find the first element >= d_next
             last = bisect.bisect_left(candidates, dnext, lo=first)
             dlen = len(d)
             candidates = {c[dlen:] for c in candidates[first:last]}
         # self._dirs includes all of the directories, recursively, so if
         # we're attempting to match foo/bar/baz.txt, it'll have '', 'foo',
         # 'foo/bar' in it. Thus we can safely ignore a candidate that has a
         # '/' in it, indicating a it's for a subdir-of-a-subdir; the
         # immediate subdir will be in there without a slash.
         ret = {c for c in candidates if b'/' not in c}
         # We really do not expect ret to be empty, since that would imply that
         # there's something in _dirs that didn't have a file in _fileset.
         assert ret
         return ret
 
     def isexact(self):
         return True
 
     @encoding.strmethod
     def __repr__(self):
         return b'<exactmatcher files=%r>' % self._files
 
 
 class differencematcher(basematcher):
     """Composes two matchers by matching if the first matches and the second
     does not.
 
     The second matcher's non-matching-attributes (bad, traversedir) are ignored.
     """
 
     def __init__(self, m1, m2):
         super(differencematcher, self).__init__()
         self._m1 = m1
         self._m2 = m2
         self.bad = m1.bad
         self.traversedir = m1.traversedir
 
     def matchfn(self, f):
         return self._m1(f) and not self._m2(f)
 
     @propertycache
     def _files(self):
         if self.isexact():
             return [f for f in self._m1.files() if self(f)]
         # If m1 is not an exact matcher, we can't easily figure out the set of
         # files, because its files() are not always files. For example, if
         # m1 is "path:dir" and m2 is "rootfileins:.", we don't
         # want to remove "dir" from the set even though it would match m2,
         # because the "dir" in m1 may not be a file.
         return self._m1.files()
 
     def visitdir(self, dir):
         if self._m2.visitdir(dir) == b'all':
             return False
         elif not self._m2.visitdir(dir):
             # m2 does not match dir, we can return 'all' here if possible
             return self._m1.visitdir(dir)
         return bool(self._m1.visitdir(dir))
 
     def visitchildrenset(self, dir):
         m2_set = self._m2.visitchildrenset(dir)
         if m2_set == b'all':
             return set()
         m1_set = self._m1.visitchildrenset(dir)
         # Possible values for m1: 'all', 'this', set(...), set()
         # Possible values for m2:        'this', set(...), set()
         # If m2 has nothing under here that we care about, return m1, even if
         # it's 'all'. This is a change in behavior from visitdir, which would
         # return True, not 'all', for some reason.
         if not m2_set:
             return m1_set
         if m1_set in [b'all', b'this']:
             # Never return 'all' here if m2_set is any kind of non-empty (either
             # 'this' or set(foo)), since m2 might return set() for a
             # subdirectory.
             return b'this'
         # Possible values for m1:         set(...), set()
         # Possible values for m2: 'this', set(...)
         # We ignore m2's set results. They're possibly incorrect:
         #  m1 = path:dir/subdir, m2=rootfilesin:dir, visitchildrenset(''):
         #    m1 returns {'dir'}, m2 returns {'dir'}, if we subtracted we'd
         #    return set(), which is *not* correct, we still need to visit 'dir'!
         return m1_set
 
     def isexact(self):
         return self._m1.isexact()
 
     @encoding.strmethod
     def __repr__(self):
         return b'<differencematcher m1=%r, m2=%r>' % (self._m1, self._m2)
 
 
 def intersectmatchers(m1, m2):
     """Composes two matchers by matching if both of them match.
 
     The second matcher's non-matching-attributes (bad, traversedir) are ignored.
     """
     if m1 is None or m2 is None:
         return m1 or m2
     if m1.always():
         m = copy.copy(m2)
         # TODO: Consider encapsulating these things in a class so there's only
         # one thing to copy from m1.
         m.bad = m1.bad
         m.traversedir = m1.traversedir
         return m
     if m2.always():
         m = copy.copy(m1)
         return m
     return intersectionmatcher(m1, m2)
 
 
 class intersectionmatcher(basematcher):
     def __init__(self, m1, m2):
         super(intersectionmatcher, self).__init__()
         self._m1 = m1
         self._m2 = m2
         self.bad = m1.bad
         self.traversedir = m1.traversedir
 
     @propertycache
     def _files(self):
         if self.isexact():
             m1, m2 = self._m1, self._m2
             if not m1.isexact():
                 m1, m2 = m2, m1
             return [f for f in m1.files() if m2(f)]
         # It neither m1 nor m2 is an exact matcher, we can't easily intersect
         # the set of files, because their files() are not always files. For
         # example, if intersecting a matcher "-I glob:foo.txt" with matcher of
         # "path:dir2", we don't want to remove "dir2" from the set.
         return self._m1.files() + self._m2.files()
 
     def matchfn(self, f):
         return self._m1(f) and self._m2(f)
 
     def visitdir(self, dir):
         visit1 = self._m1.visitdir(dir)
         if visit1 == b'all':
             return self._m2.visitdir(dir)
         # bool() because visit1=True + visit2='all' should not be 'all'
         return bool(visit1 and self._m2.visitdir(dir))
 
     def visitchildrenset(self, dir):
         m1_set = self._m1.visitchildrenset(dir)
         if not m1_set:
             return set()
         m2_set = self._m2.visitchildrenset(dir)
         if not m2_set:
             return set()
 
         if m1_set == b'all':
             return m2_set
         elif m2_set == b'all':
             return m1_set
 
         if m1_set == b'this' or m2_set == b'this':
             return b'this'
 
         assert isinstance(m1_set, set) and isinstance(m2_set, set)
         return m1_set.intersection(m2_set)
 
     def always(self):
         return self._m1.always() and self._m2.always()
 
     def isexact(self):
         return self._m1.isexact() or self._m2.isexact()
 
     @encoding.strmethod
     def __repr__(self):
         return b'<intersectionmatcher m1=%r, m2=%r>' % (self._m1, self._m2)
 
 
 class subdirmatcher(basematcher):
     """Adapt a matcher to work on a subdirectory only.
 
     The paths are remapped to remove/insert the path as needed:
 
     >>> from . import pycompat
     >>> m1 = match(util.localpath(b'/root'), b'', [b'a.txt', b'sub/b.txt'], auditor=lambda name: None)
     >>> m2 = subdirmatcher(b'sub', m1)
     >>> m2(b'a.txt')
     False
     >>> m2(b'b.txt')
     True
     >>> m2.matchfn(b'a.txt')
     False
     >>> m2.matchfn(b'b.txt')
     True
     >>> m2.files()
     ['b.txt']
     >>> m2.exact(b'b.txt')
     True
     >>> def bad(f, msg):
     ...     print(pycompat.sysstr(b"%s: %s" % (f, msg)))
     >>> m1.bad = bad
     >>> m2.bad(b'x.txt', b'No such file')
     sub/x.txt: No such file
     """
 
     def __init__(self, path, matcher):
         super(subdirmatcher, self).__init__()
         self._path = path
         self._matcher = matcher
         self._always = matcher.always()
 
         self._files = [
             f[len(path) + 1 :]
             for f in matcher._files
             if f.startswith(path + b"/")
         ]
 
         # If the parent repo had a path to this subrepo and the matcher is
         # a prefix matcher, this submatcher always matches.
         if matcher.prefix():
             self._always = any(f == path for f in matcher._files)
 
     def bad(self, f, msg):
         self._matcher.bad(self._path + b"/" + f, msg)
 
     def matchfn(self, f):
         # Some information is lost in the superclass's constructor, so we
         # can not accurately create the matching function for the subdirectory
         # from the inputs. Instead, we override matchfn() and visitdir() to
         # call the original matcher with the subdirectory path prepended.
         return self._matcher.matchfn(self._path + b"/" + f)
 
     def visitdir(self, dir):
         if dir == b'':
             dir = self._path
         else:
             dir = self._path + b"/" + dir
         return self._matcher.visitdir(dir)
 
     def visitchildrenset(self, dir):
         if dir == b'':
             dir = self._path
         else:
             dir = self._path + b"/" + dir
         return self._matcher.visitchildrenset(dir)
 
     def always(self):
         return self._always
 
     def prefix(self):
         return self._matcher.prefix() and not self._always
 
     @encoding.strmethod
     def __repr__(self):
         return b'<subdirmatcher path=%r, matcher=%r>' % (
             self._path,
             self._matcher,
         )
 
 
 class prefixdirmatcher(basematcher):
     """Adapt a matcher to work on a parent directory.
 
     The matcher's non-matching-attributes (bad, traversedir) are ignored.
 
     The prefix path should usually be the relative path from the root of
     this matcher to the root of the wrapped matcher.
 
     >>> m1 = match(util.localpath(b'/root/d/e'), b'f', [b'../a.txt', b'b.txt'], auditor=lambda name: None)
     >>> m2 = prefixdirmatcher(b'd/e', m1)
     >>> m2(b'a.txt')
     False
     >>> m2(b'd/e/a.txt')
     True
     >>> m2(b'd/e/b.txt')
     False
     >>> m2.files()
     ['d/e/a.txt', 'd/e/f/b.txt']
     >>> m2.exact(b'd/e/a.txt')
     True
     >>> m2.visitdir(b'd')
     True
     >>> m2.visitdir(b'd/e')
     True
     >>> m2.visitdir(b'd/e/f')
     True
     >>> m2.visitdir(b'd/e/g')
     False
     >>> m2.visitdir(b'd/ef')
     False
     """
 
     def __init__(self, path, matcher, badfn=None):
         super(prefixdirmatcher, self).__init__(badfn)
         if not path:
             raise error.ProgrammingError(b'prefix path must not be empty')
         self._path = path
         self._pathprefix = path + b'/'
         self._matcher = matcher
 
     @propertycache
     def _files(self):
         return [self._pathprefix + f for f in self._matcher._files]
 
     def matchfn(self, f):
         if not f.startswith(self._pathprefix):
             return False
         return self._matcher.matchfn(f[len(self._pathprefix) :])
 
     @propertycache
     def _pathdirs(self):
         return set(pathutil.finddirs(self._path))
 
     def visitdir(self, dir):
         if dir == self._path:
             return self._matcher.visitdir(b'')
         if dir.startswith(self._pathprefix):
             return self._matcher.visitdir(dir[len(self._pathprefix) :])
         return dir in self._pathdirs
 
     def visitchildrenset(self, dir):
         if dir == self._path:
             return self._matcher.visitchildrenset(b'')
         if dir.startswith(self._pathprefix):
             return self._matcher.visitchildrenset(dir[len(self._pathprefix) :])
         if dir in self._pathdirs:
             return b'this'
         return set()
 
     def isexact(self):
         return self._matcher.isexact()
 
     def prefix(self):
         return self._matcher.prefix()
 
     @encoding.strmethod
     def __repr__(self):
         return b'<prefixdirmatcher path=%r, matcher=%r>' % (
             pycompat.bytestr(self._path),
             self._matcher,
         )
 
 
 class unionmatcher(basematcher):
     """A matcher that is the union of several matchers.
 
     The non-matching-attributes (bad, traversedir) are taken from the first
     matcher.
     """
 
     def __init__(self, matchers):
         m1 = matchers[0]
         super(unionmatcher, self).__init__()
         self.traversedir = m1.traversedir
         self._matchers = matchers
 
     def matchfn(self, f):
         for match in self._matchers:
             if match(f):
                 return True
         return False
 
     def visitdir(self, dir):
         r = False
         for m in self._matchers:
             v = m.visitdir(dir)
             if v == b'all':
                 return v
             r |= v
         return r
 
     def visitchildrenset(self, dir):
         r = set()
         this = False
         for m in self._matchers:
             v = m.visitchildrenset(dir)
             if not v:
                 continue
             if v == b'all':
                 return v
             if this or v == b'this':
                 this = True
                 # don't break, we might have an 'all' in here.
                 continue
             assert isinstance(v, set)
             r = r.union(v)
         if this:
             return b'this'
         return r
 
     @encoding.strmethod
     def __repr__(self):
         return b'<unionmatcher matchers=%r>' % self._matchers
 
 
 def patkind(pattern, default=None):
     r"""If pattern is 'kind:pat' with a known kind, return kind.
 
     >>> patkind(br're:.*\.c$')
     're'
     >>> patkind(b'glob:*.c')
     'glob'
     >>> patkind(b'relpath:test.py')
     'relpath'
     >>> patkind(b'main.py')
     >>> patkind(b'main.py', default=b're')
     're'
     """
     return _patsplit(pattern, default)[0]
 
 
 def _patsplit(pattern, default):
     """Split a string into the optional pattern kind prefix and the actual
     pattern."""
     if b':' in pattern:
         kind, pat = pattern.split(b':', 1)
         if kind in allpatternkinds:
             return kind, pat
     return default, pattern
 
 
 def _globre(pat):
     r"""Convert an extended glob string to a regexp string.
 
     >>> from . import pycompat
     >>> def bprint(s):
     ...     print(pycompat.sysstr(s))
     >>> bprint(_globre(br'?'))
     .
     >>> bprint(_globre(br'*'))
     [^/]*
     >>> bprint(_globre(br'**'))
     .*
     >>> bprint(_globre(br'**/a'))
     (?:.*/)?a
     >>> bprint(_globre(br'a/**/b'))
     a/(?:.*/)?b
     >>> bprint(_globre(br'[a*?!^][^b][!c]'))
     [a*?!^][\^b][^c]
     >>> bprint(_globre(br'{a,b}'))
     (?:a|b)
     >>> bprint(_globre(br'.\*\?'))
     \.\*\?
     """
     i, n = 0, len(pat)
     res = b''
     group = 0
     escape = util.stringutil.regexbytesescapemap.get
 
     def peek():
         return i < n and pat[i : i + 1]
 
     while i < n:
         c = pat[i : i + 1]
         i += 1
         if c not in b'*?[{},\\':
             res += escape(c, c)
         elif c == b'*':
             if peek() == b'*':
                 i += 1
                 if peek() == b'/':
                     i += 1
                     res += b'(?:.*/)?'
                 else:
                     res += b'.*'
             else:
                 res += b'[^/]*'
         elif c == b'?':
             res += b'.'
         elif c == b'[':
             j = i
             if j < n and pat[j : j + 1] in b'!]':
                 j += 1
             while j < n and pat[j : j + 1] != b']':
                 j += 1
             if j >= n:
                 res += b'\\['
             else:
                 stuff = pat[i:j].replace(b'\\', b'\\\\')
                 i = j + 1
                 if stuff[0:1] == b'!':
                     stuff = b'^' + stuff[1:]
                 elif stuff[0:1] == b'^':
                     stuff = b'\\' + stuff
                 res = b'%s[%s]' % (res, stuff)
         elif c == b'{':
             group += 1
             res += b'(?:'
         elif c == b'}' and group:
             res += b')'
             group -= 1
         elif c == b',' and group:
             res += b'|'
         elif c == b'\\':
             p = peek()
             if p:
                 i += 1
                 res += escape(p, p)
             else:
                 res += escape(c, c)
         else:
             res += escape(c, c)
     return res
 
 
 FLAG_RE = util.re.compile(br'^\(\?([aiLmsux]+)\)(.*)')
 
 
 def _regex(kind, pat, globsuffix):
     """Convert a (normalized) pattern of any kind into a
     regular expression.
     globsuffix is appended to the regexp of globs."""
     if not pat and kind in (b'glob', b'relpath'):
         return b''
     if kind == b're':
         return pat
     if kind == b'filepath':
         raise error.ProgrammingError(
             "'filepath:' patterns should not be converted to a regex"
         )
     if kind in (b'path', b'relpath'):
         if pat == b'.':
             return b''
         return util.stringutil.reescape(pat) + b'(?:/|$)'
     if kind == b'rootfilesin':
         if pat == b'.':
             escaped = b''
         else:
             # Pattern is a directory name.
             escaped = util.stringutil.reescape(pat) + b'/'
         # Anything after the pattern must be a non-directory.
         return escaped + b'[^/]+$'
     if kind == b'relglob':
         globre = _globre(pat)
         if globre.startswith(b'[^/]*'):
             # When pat has the form *XYZ (common), make the returned regex more
             # legible by returning the regex for **XYZ instead of **/*XYZ.
             return b'.*' + globre[len(b'[^/]*') :] + globsuffix
         return b'(?:|.*/)' + globre + globsuffix
     if kind == b'relre':
         flag = None
         m = FLAG_RE.match(pat)
         if m:
             flag, pat = m.groups()
         if not pat.startswith(b'^'):
             pat = b'.*' + pat
         if flag is not None:
             pat = br'(?%s:%s)' % (flag, pat)
         return pat
     if kind in (b'glob', b'rootglob'):
         return _globre(pat) + globsuffix
     raise error.ProgrammingError(b'not a regex pattern: %s:%s' % (kind, pat))
 
 
 def _buildmatch(kindpats, globsuffix, root):
     """Return regexp string and a matcher function for kindpats.
     globsuffix is appended to the regexp of globs."""
     matchfuncs = []
 
     subincludes, kindpats = _expandsubinclude(kindpats, root)
     if subincludes:
         submatchers = {}
 
         def matchsubinclude(f):
             for prefix, matcherargs in subincludes:
                 if f.startswith(prefix):
                     mf = submatchers.get(prefix)
                     if mf is None:
                         mf = match(*matcherargs)
                         submatchers[prefix] = mf
 
                     if mf(f[len(prefix) :]):
                         return True
             return False
 
         matchfuncs.append(matchsubinclude)
 
     regex = b''
     if kindpats:
         if all(k == b'rootfilesin' for k, p, s in kindpats):
             dirs = {p for k, p, s in kindpats}
 
             def mf(f):
                 i = f.rfind(b'/')
                 if i >= 0:
                     dir = f[:i]
                 else:
                     dir = b'.'
                 return dir in dirs
 
             regex = b'rootfilesin: %s' % stringutil.pprint(list(sorted(dirs)))
             matchfuncs.append(mf)
         else:
             regex, mf = _buildregexmatch(kindpats, globsuffix)
             matchfuncs.append(mf)
 
     if len(matchfuncs) == 1:
         return regex, matchfuncs[0]
     else:
         return regex, lambda f: any(mf(f) for mf in matchfuncs)
 
 
 MAX_RE_SIZE = 20000
 
 
 def _joinregexes(regexps):
     """gather multiple regular expressions into a single one"""
     return b'|'.join(regexps)
 
 
 def _buildregexmatch(kindpats, globsuffix):
     """Build a match function from a list of kinds and kindpats,
     return regexp string and a matcher function.
 
     Test too large input
     >>> _buildregexmatch([
     ...     (b'relglob', b'?' * MAX_RE_SIZE, b'')
     ... ], b'$')
     Traceback (most recent call last):
     ...
     Abort: matcher pattern is too long (20009 bytes)
     """
     try:
         allgroups = []
         regexps = []
         exact = set()
         for kind, pattern, _source in kindpats:
             if kind == b'filepath':
                 exact.add(pattern)
                 continue
             regexps.append(_regex(kind, pattern, globsuffix))
 
         fullregexp = _joinregexes(regexps)
 
         startidx = 0
         groupsize = 0
         for idx, r in enumerate(regexps):
             piecesize = len(r)
             if piecesize > MAX_RE_SIZE:
                 msg = _(b"matcher pattern is too long (%d bytes)") % piecesize
                 raise error.Abort(msg)
             elif (groupsize + piecesize) > MAX_RE_SIZE:
                 group = regexps[startidx:idx]
                 allgroups.append(_joinregexes(group))
                 startidx = idx
                 groupsize = 0
             groupsize += piecesize + 1
 
         if startidx == 0:
             matcher = _rematcher(fullregexp)
             func = lambda s: bool(matcher(s))
         else:
             group = regexps[startidx:]
             allgroups.append(_joinregexes(group))
             allmatchers = [_rematcher(g) for g in allgroups]
             func = lambda s: any(m(s) for m in allmatchers)
 
         actualfunc = func
         if exact:
             # An empty regex will always match, so only call the regex if
             # there were any actual patterns to match.
             if not regexps:
                 actualfunc = lambda s: s in exact
             else:
                 actualfunc = lambda s: s in exact or func(s)
         return fullregexp, actualfunc
     except re.error:
         for k, p, s in kindpats:
             if k == b'filepath':
                 continue
             try:
                 _rematcher(_regex(k, p, globsuffix))
             except re.error:
                 if s:
                     raise error.Abort(
                         _(b"%s: invalid pattern (%s): %s") % (s, k, p)
                     )
                 else:
                     raise error.Abort(_(b"invalid pattern (%s): %s") % (k, p))
         raise error.Abort(_(b"invalid pattern"))
 
 
 def _patternrootsanddirs(kindpats):
     """Returns roots and directories corresponding to each pattern.
 
     This calculates the roots and directories exactly matching the patterns and
     returns a tuple of (roots, dirs) for each. It does not return other
     directories which may also need to be considered, like the parent
     directories.
     """
     r = []
     d = []
     for kind, pat, source in kindpats:
         if kind in (b'glob', b'rootglob'):  # find the non-glob prefix
             root = []
             for p in pat.split(b'/'):
                 if b'[' in p or b'{' in p or b'*' in p or b'?' in p:
                     break
                 root.append(p)
             r.append(b'/'.join(root))
         elif kind in (b'relpath', b'path', b'filepath'):
             if pat == b'.':
                 pat = b''
             r.append(pat)
         elif kind in (b'rootfilesin',):
             if pat == b'.':
                 pat = b''
             d.append(pat)
         else:  # relglob, re, relre
             r.append(b'')
     return r, d
 
 
 def _roots(kindpats):
     '''Returns root directories to match recursively from the given patterns.'''
     roots, dirs = _patternrootsanddirs(kindpats)
     return roots
 
 
 def _rootsdirsandparents(kindpats):
     """Returns roots and exact directories from patterns.
 
     `roots` are directories to match recursively, `dirs` should
     be matched non-recursively, and `parents` are the implicitly required
     directories to walk to items in either roots or dirs.
 
     Returns a tuple of (roots, dirs, parents).
 
     >>> r = _rootsdirsandparents(
     ...     [(b'glob', b'g/h/*', b''), (b'glob', b'g/h', b''),
     ...      (b'glob', b'g*', b'')])
     >>> print(r[0:2], sorted(r[2])) # the set has an unstable output
     (['g/h', 'g/h', ''], []) ['', 'g']
     >>> r = _rootsdirsandparents(
     ...     [(b'rootfilesin', b'g/h', b''), (b'rootfilesin', b'', b'')])
     >>> print(r[0:2], sorted(r[2])) # the set has an unstable output
     ([], ['g/h', '']) ['', 'g']
     >>> r = _rootsdirsandparents(
     ...     [(b'relpath', b'r', b''), (b'path', b'p/p', b''),
     ...      (b'path', b'', b'')])
     >>> print(r[0:2], sorted(r[2])) # the set has an unstable output
     (['r', 'p/p', ''], []) ['', 'p']
     >>> r = _rootsdirsandparents(
     ...     [(b'relglob', b'rg*', b''), (b're', b're/', b''),
     ...      (b'relre', b'rr', b'')])
     >>> print(r[0:2], sorted(r[2])) # the set has an unstable output
     (['', '', ''], []) ['']
     """
     r, d = _patternrootsanddirs(kindpats)
 
     p = set()
     # Add the parents as non-recursive/exact directories, since they must be
     # scanned to get to either the roots or the other exact directories.
     p.update(pathutil.dirs(d))
     p.update(pathutil.dirs(r))
 
     # FIXME: all uses of this function convert these to sets, do so before
     # returning.
     # FIXME: all uses of this function do not need anything in 'roots' and
     # 'dirs' to also be in 'parents', consider removing them before returning.
     return r, d, p
 
 
 def _explicitfiles(kindpats):
     """Returns the potential explicit filenames from the patterns.
 
     >>> _explicitfiles([(b'path', b'foo/bar', b'')])
     ['foo/bar']
     >>> _explicitfiles([(b'rootfilesin', b'foo/bar', b'')])
     []
     """
     # Keep only the pattern kinds where one can specify filenames (vs only
     # directory names).
     filable = [kp for kp in kindpats if kp[0] not in (b'rootfilesin',)]
     return _roots(filable)
 
 
 def _prefix(kindpats):
     '''Whether all the patterns match a prefix (i.e. recursively)'''
     for kind, pat, source in kindpats:
         if kind not in (b'path', b'relpath'):
             return False
     return True
 
 
 _commentre = None
 
 
 def readpatternfile(filepath, warn, sourceinfo=False):
     """parse a pattern file, returning a list of
     patterns. These patterns should be given to compile()
     to be validated and converted into a match function.
 
     trailing white space is dropped.
     the escape character is backslash.
     comments start with #.
     empty lines are skipped.
 
     lines can be of the following formats:
 
     syntax: regexp # defaults following lines to non-rooted regexps
     syntax: glob   # defaults following lines to non-rooted globs
     re:pattern     # non-rooted regular expression
     glob:pattern   # non-rooted glob
     rootglob:pat   # rooted glob (same root as ^ in regexps)
     pattern        # pattern of the current default type
 
     if sourceinfo is set, returns a list of tuples:
     (pattern, lineno, originalline).
     This is useful to debug ignore patterns.
     """
 
     syntaxes = {
         b're': b'relre:',
         b'regexp': b'relre:',
         b'glob': b'relglob:',
         b'rootglob': b'rootglob:',
         b'include': b'include',
         b'subinclude': b'subinclude',
     }
     syntax = b'relre:'
     patterns = []
 
     fp = open(filepath, b'rb')
     for lineno, line in enumerate(fp, start=1):
         if b"#" in line:
             global _commentre
             if not _commentre:
                 _commentre = util.re.compile(br'((?:^|[^\\])(?:\\\\)*)#.*')
             # remove comments prefixed by an even number of escapes
             m = _commentre.search(line)
             if m:
                 line = line[: m.end(1)]
             # fixup properly escaped comments that survived the above
             line = line.replace(b"\\#", b"#")
         line = line.rstrip()
         if not line:
             continue
 
         if line.startswith(b'syntax:'):
             s = line[7:].strip()
             try:
                 syntax = syntaxes[s]
             except KeyError:
                 if warn:
                     warn(
                         _(b"%s: ignoring invalid syntax '%s'\n") % (filepath, s)
                     )
             continue
 
         linesyntax = syntax
         for s, rels in syntaxes.items():
             if line.startswith(rels):
                 linesyntax = rels
                 line = line[len(rels) :]
                 break
             elif line.startswith(s + b':'):
                 linesyntax = rels
                 line = line[len(s) + 1 :]
                 break
         if sourceinfo:
             patterns.append((linesyntax + line, lineno, line))
         else:
             patterns.append(linesyntax + line)
     fp.close()
     return patterns