subrepo: fix a crash when archiving an svn or git subrepo

Only hgsubrepos have a repository attribute.  This is pretty hacky, but probably
the best we can do on stable.  Pushing the lfstatus check down into the wrapper
for hgsubrepo (and dropping the check for lfstatus at the top of
`hgsubrepoarchive()`) resulted in various test failures because:

  1) hgsubrepoarchive isn't returning the number of files archived at the
     bottom, resulting in an error about += NoneType
  2) These copypasta archive wrappers don't use progress bars
  3) Largefiles are *not* currently archived when using extdiff (68822b7cdd01),
     but pushing this context manager down into the subrepo resulted in it
     apparently doing so (as evidenced by progress bars being dropped)

The other uses of `lfstatus()` are not in the substate processing loop, so they
shouldn't be an issue.

I initially put testcases in this test for largefiles-{on,off}, and it flagged
a bunch of exit code differences for `cat` and `diff`, so I backed that off.

Differential Revision: https://phab.mercurial-scm.org/D7714

hgext/largefiles/overrides.py

@@ -1,1808 +1,1825 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'''
 from __future__ import absolute_import
 
 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,
     error,
     exchange,
     extensions,
     exthelper,
     filemerge,
     hg,
     logcmdutil,
     match as matchmod,
     merge,
     pathutil,
     pycompat,
     scmutil,
     smartset,
     subrepo,
     upgrade,
     url as urlmod,
     util,
 )
 
 from . import (
     lfcommands,
     lfutil,
     storefactory,
 )
 
 eh = exthelper.exthelper()
 
 # -- Utility functions: commonly/repeatedly needed functionality ---------------
 
 
 def composelargefilematcher(match, manifest):
     '''create a matcher that matches only the largefiles in the original
     matcher'''
     m = copy.copy(match)
     lfile = lambda f: lfutil.standin(f) in manifest
     m._files = [lf for lf in m._files if lfile(lf)]
     m._fileset = set(m._files)
     m.always = lambda: False
     origmatchfn = m.matchfn
     m.matchfn = lambda f: lfile(f) and origmatchfn(f)
     return m
 
 
 def composenormalfilematcher(match, manifest, exclude=None):
     excluded = set()
     if exclude is not None:
         excluded.update(exclude)
 
     m = copy.copy(match)
     notlfile = lambda f: not (
         lfutil.isstandin(f) or lfutil.standin(f) in manifest or f in excluded
     )
     m._files = [lf for lf in m._files if notlfile(lf)]
     m._fileset = set(m._files)
     m.always = lambda: False
     origmatchfn = m.matchfn
     m.matchfn = lambda f: notlfile(f) and origmatchfn(f)
     return m
 
 
 def addlargefiles(ui, repo, isaddremove, matcher, uipathfn, **opts):
     large = opts.get(r'large')
     lfsize = lfutil.getminsize(
         ui, lfutil.islfilesrepo(repo), opts.get(r'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(r'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)
                 if lfdirstate[f] == b'r':
                     lfdirstate.normallookup(f)
                 else:
                     lfdirstate.add(f)
             lfdirstate.write()
             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
 
 
 @contextlib.contextmanager
 def lfstatus(repo):
     oldvalue = getattr(repo, 'lfstatus', False)
     repo.lfstatus = True
     try:
         yield
     finally:
         repo.lfstatus = oldvalue
 
 
 def removelargefiles(ui, repo, isaddremove, matcher, uipathfn, dryrun, **opts):
     after = opts.get(r'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:
             lfutil.synclfdirstate(
                 repo, lfdirstate, lfutil.splitstandin(f), False
             )
 
         lfdirstate.write()
 
     return result
 
 
 # For overriding mercurial.hgweb.webcommands so that largefiles will
 # appear at their right place in the manifests.
 @eh.wrapfunction(webcommands, b'decodepath')
 def decodepath(orig, path):
     return lfutil.splitstandin(path) or path
 
 
 # -- Wrappers: modify existing commands --------------------------------
 
 
 @eh.wrapcommand(
     b'add',
     opts=[
         (b'', b'large', None, _(b'add as largefile')),
         (b'', b'normal', None, _(b'add as normal file')),
         (
             b'',
             b'lfsize',
             b'',
             _(
                 b'add all files above this size (in megabytes) '
                 b'as largefiles (default: 10)'
             ),
         ),
     ],
 )
 def overrideadd(orig, ui, repo, *pats, **opts):
     if opts.get(r'normal') and opts.get(r'large'):
         raise error.Abort(_(b'--normal cannot be used with --large'))
     return orig(ui, repo, *pats, **opts)
 
 
 @eh.wrapfunction(cmdutil, b'add')
 def cmdutiladd(orig, ui, repo, matcher, prefix, uipathfn, explicitonly, **opts):
     # The --normal flag short circuits this override
     if opts.get(r'normal'):
         return orig(ui, repo, matcher, prefix, uipathfn, explicitonly, **opts)
 
     ladded, lbad = addlargefiles(ui, repo, False, matcher, uipathfn, **opts)
     normalmatcher = composenormalfilematcher(
         matcher, repo[None].manifest(), ladded
     )
     bad = orig(ui, repo, normalmatcher, prefix, uipathfn, explicitonly, **opts)
 
     bad.extend(f for f in lbad)
     return bad
 
 
 @eh.wrapfunction(cmdutil, b'remove')
 def cmdutilremove(
     orig, ui, repo, matcher, prefix, uipathfn, after, force, subrepos, dryrun
 ):
     normalmatcher = composenormalfilematcher(matcher, repo[None].manifest())
     result = orig(
         ui,
         repo,
         normalmatcher,
         prefix,
         uipathfn,
         after,
         force,
         subrepos,
         dryrun,
     )
     return (
         removelargefiles(
             ui, repo, False, matcher, uipathfn, dryrun, after=after, force=force
         )
         or result
     )
 
 
 @eh.wrapfunction(subrepo.hgsubrepo, b'status')
 def overridestatusfn(orig, repo, rev2, **opts):
     with lfstatus(repo._repo):
         return orig(repo, rev2, **opts)
 
 
 @eh.wrapcommand(b'status')
 def overridestatus(orig, ui, repo, *pats, **opts):
     with lfstatus(repo):
         return orig(ui, repo, *pats, **opts)
 
 
 @eh.wrapfunction(subrepo.hgsubrepo, b'dirty')
 def overridedirty(orig, repo, ignoreupdate=False, missing=False):
     with lfstatus(repo._repo):
         return orig(repo, ignoreupdate=ignoreupdate, missing=missing)
 
 
 @eh.wrapcommand(b'log')
 def overridelog(orig, ui, repo, *pats, **opts):
     def overridematchandpats(
         orig,
         ctx,
         pats=(),
         opts=None,
         globbed=False,
         default=b'relpath',
         badfn=None,
     ):
         """Matcher that merges root directory with .hglf, suitable for log.
         It is still possible to match .hglf directly.
         For any listed files run log on the standin too.
         matchfn tries both the given filename and with .hglf stripped.
         """
         if opts is None:
             opts = {}
         matchandpats = orig(ctx, pats, opts, globbed, default, badfn=badfn)
         m, p = copy.copy(matchandpats)
 
         if m.always():
             # We want to match everything anyway, so there's no benefit trying
             # to add standins.
             return matchandpats
 
         pats = set(p)
 
         def fixpats(pat, tostandin=lfutil.standin):
             if pat.startswith(b'set:'):
                 return pat
 
             kindpat = matchmod._patsplit(pat, None)
 
             if kindpat[0] is not None:
                 return kindpat[0] + b':' + tostandin(kindpat[1])
             return tostandin(kindpat[1])
 
         cwd = repo.getcwd()
         if cwd:
             hglf = lfutil.shortname
             back = util.pconvert(repo.pathto(hglf)[: -len(hglf)])
 
             def tostandin(f):
                 # The file may already be a standin, so truncate the back
                 # prefix and test before mangling it.  This avoids turning
                 # 'glob:../.hglf/foo*' into 'glob:../.hglf/../.hglf/foo*'.
                 if f.startswith(back) and lfutil.splitstandin(f[len(back) :]):
                     return f
 
                 # An absolute path is from outside the repo, so truncate the
                 # path to the root before building the standin.  Otherwise cwd
                 # is somewhere in the repo, relative to root, and needs to be
                 # prepended before building the standin.
                 if os.path.isabs(cwd):
                     f = f[len(back) :]
                 else:
                     f = cwd + b'/' + f
                 return back + lfutil.standin(f)
 
         else:
 
             def tostandin(f):
                 if lfutil.isstandin(f):
                     return f
                 return lfutil.standin(f)
 
         pats.update(fixpats(f, tostandin) for f in p)
 
         for i in range(0, len(m._files)):
             # Don't add '.hglf' to m.files, since that is already covered by '.'
             if m._files[i] == b'.':
                 continue
             standin = lfutil.standin(m._files[i])
             # If the "standin" is a directory, append instead of replace to
             # support naming a directory on the command line with only
             # largefiles.  The original directory is kept to support normal
             # files.
             if standin in ctx:
                 m._files[i] = standin
             elif m._files[i] not in ctx and repo.wvfs.isdir(standin):
                 m._files.append(standin)
 
         m._fileset = set(m._files)
         m.always = lambda: False
         origmatchfn = m.matchfn
 
         def lfmatchfn(f):
             lf = lfutil.splitstandin(f)
             if lf is not None and origmatchfn(lf):
                 return True
             r = origmatchfn(f)
             return r
 
         m.matchfn = lfmatchfn
 
         ui.debug(b'updated patterns: %s\n' % b', '.join(sorted(pats)))
         return m, pats
 
     # For hg log --patch, the match object is used in two different senses:
     # (1) to determine what revisions should be printed out, and
     # (2) to determine what files to print out diffs for.
     # The magic matchandpats override should be used for case (1) but not for
     # case (2).
     oldmatchandpats = scmutil.matchandpats
 
     def overridemakefilematcher(orig, repo, pats, opts, badfn=None):
         wctx = repo[None]
         match, pats = oldmatchandpats(wctx, pats, opts, badfn=badfn)
         return lambda ctx: match
 
     wrappedmatchandpats = extensions.wrappedfunction(
         scmutil, b'matchandpats', overridematchandpats
     )
     wrappedmakefilematcher = extensions.wrappedfunction(
         logcmdutil, b'_makenofollowfilematcher', overridemakefilematcher
     )
     with wrappedmatchandpats, wrappedmakefilematcher:
         return orig(ui, repo, *pats, **opts)
 
 
 @eh.wrapcommand(
     b'verify',
     opts=[
         (
             b'',
             b'large',
             None,
             _(b'verify that all largefiles in current revision exists'),
         ),
         (
             b'',
             b'lfa',
             None,
             _(b'verify largefiles in all revisions, not just current'),
         ),
         (
             b'',
             b'lfc',
             None,
             _(b'verify local largefile contents, not just existence'),
         ),
     ],
 )
 def overrideverify(orig, ui, repo, *pats, **opts):
     large = opts.pop(r'large', False)
     all = opts.pop(r'lfa', False)
     contents = opts.pop(r'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(r'large', False)
     if large:
 
         class fakerepo(object):
             dirstate = lfutil.openlfdirstate(ui, repo)
 
         orig(ui, fakerepo, *pats, **opts)
     else:
         orig(ui, repo, *pats, **opts)
 
 
 # Before starting the manifest merge, merge.updates will call
 # _checkunknownfile to check if there are any files in the merged-in
 # changeset that collide with unknown files in the working copy.
 #
 # The largefiles are seen as unknown, so this prevents us from merging
 # in a file 'foo' if we already have a largefile with the same name.
 #
 # The overridden function filters the unknown files by removing any
 # largefiles. This makes the merge proceed and we can then handle this
 # case further in the overridden calculateupdates function below.
 @eh.wrapfunction(merge, b'_checkunknownfile')
 def overridecheckunknownfile(origfn, repo, wctx, mctx, f, f2=None):
     if lfutil.standin(repo.dirstate.normalize(f)) in wctx:
         return False
     return origfn(repo, wctx, mctx, f, f2)
 
 
 # The manifest merge handles conflicts on the manifest level. We want
 # to handle changes in largefile-ness of files at this level too.
 #
 # The strategy is to run the original calculateupdates and then process
 # the action list it outputs. There are two cases we need to deal with:
 #
 # 1. Normal file in p1, largefile in p2. Here the largefile is
 #    detected via its standin file, which will enter the working copy
 #    with a "get" action. It is not "merge" since the standin is all
 #    Mercurial is concerned with at this level -- the link to the
 #    existing normal file is not relevant here.
 #
 # 2. Largefile in p1, normal file in p2. Here we get a "merge" action
 #    since the largefile will be present in the working copy and
 #    different from the normal file in p2. Mercurial therefore
 #    triggers a merge action.
 #
 # In both cases, we prompt the user and emit new actions to either
 # remove the standin (if the normal file was kept) or to remove the
 # normal file and get the standin (if the largefile was kept). The
 # default prompt answer is to use the largefile version since it was
 # presumably changed on purpose.
 #
 # Finally, the merge.applyupdates function will then take care of
 # writing the files into the working copy and lfcommands.updatelfiles
 # will update the largefiles.
 @eh.wrapfunction(merge, b'calculateupdates')
 def overridecalculateupdates(
     origfn, repo, p1, p2, pas, branchmerge, force, acceptremote, *args, **kwargs
 ):
     overwrite = force and not branchmerge
     actions, diverge, renamedelete = origfn(
         repo, p1, p2, pas, branchmerge, force, acceptremote, *args, **kwargs
     )
 
     if overwrite:
         return actions, diverge, renamedelete
 
     # Convert to dictionary with filename as key and action as value.
     lfiles = set()
     for f in actions:
         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) = actions.get(lfile, (None, None, None))
         (sm, sargs, smsg) = actions.get(standin, (None, None, None))
         if sm in (b'g', b'dc') and lm != b'r':
             if sm == b'dc':
                 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
                 actions[lfile] = (b'r', None, b'replaced by standin')
                 actions[standin] = (b'g', sargs, b'replaces standin')
             else:  # keep local normal file
                 actions[lfile] = (b'k', None, b'replaces standin')
                 if branchmerge:
                     actions[standin] = (b'k', None, b'replaced by non-standin')
                 else:
                     actions[standin] = (b'r', None, b'replaced by non-standin')
         elif lm in (b'g', b'dc') and sm != b'r':
             if lm == b'dc':
                 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
                     actions[lfile] = (b'k', None, b'replaced by standin')
                     actions[standin] = (b'k', None, b'replaces standin')
                 else:
                     # "lfile" should be marked as "removed" without
                     # removal of itself
                     actions[lfile] = (
                         b'lfmr',
                         None,
                         b'forget non-standin largefile',
                     )
 
                     # linear-merge should treat this largefile as 're-added'
                     actions[standin] = (b'a', None, b'keep standin')
             else:  # pick remote normal file
                 actions[lfile] = (b'g', largs, b'replaces standin')
                 actions[standin] = (b'r', None, b'replaced by non-standin')
 
     return actions, diverge, renamedelete
 
 
 @eh.wrapfunction(merge, b'recordupdates')
 def mergerecordupdates(orig, repo, actions, branchmerge, getfiledata):
     if b'lfmr' in actions:
         lfdirstate = lfutil.openlfdirstate(repo.ui, repo)
         for lfile, args, msg in actions[b'lfmr']:
             # this should be executed before 'orig', to execute 'remove'
             # before all other actions
             repo.dirstate.remove(lfile)
             # make sure lfile doesn't get synclfdirstate'd as normal
             lfdirstate.add(lfile)
         lfdirstate.write()
 
     return orig(repo, actions, branchmerge, getfiledata)
 
 
 # Override filemerge to prompt the user about how they wish to merge
 # largefiles. This will handle identical edits without prompting the user.
 @eh.wrapfunction(filemerge, b'_filemerge')
 def overridefilemerge(
     origfn, premerge, repo, wctx, mynode, orig, fcd, fco, fca, labels=None
 ):
     if not lfutil.isstandin(orig) or fcd.isabsent() or fco.isabsent():
         return origfn(
             premerge, 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 True, 0, False
 
 
 @eh.wrapfunction(copiesmod, b'pathcopies')
 def copiespathcopies(orig, ctx1, ctx2, match=None):
     copies = orig(ctx1, ctx2, match=match)
     updated = {}
 
     for k, v in pycompat.iteritems(copies):
         updated[lfutil.splitstandin(k) or k] = lfutil.splitstandin(v) or v
 
     return updated
 
 
 # Copy first changes the matchers to match standins instead of
 # largefiles.  Then it overrides util.copyfile in that function it
 # checks if the destination largefile already exists. It also keeps a
 # list of copied files so that the largefiles can be copied and the
 # dirstate updated.
 @eh.wrapfunction(cmdutil, b'copy')
 def overridecopy(orig, ui, repo, pats, opts, rename=False):
     # doesn't remove largefile on rename
     if len(pats) < 2:
         # this isn't legal, let the original function deal with it
         return orig(ui, repo, pats, opts, rename)
 
     # This could copy both lfiles and normal files in one command,
     # but we don't want to do that. First replace their matcher to
     # only match normal files and run it, then replace it to just
     # match largefiles and run it again.
     nonormalfiles = False
     nolfiles = False
     manifest = repo[None].manifest()
 
     def normalfilesmatchfn(
         orig,
         ctx,
         pats=(),
         opts=None,
         globbed=False,
         default=b'relpath',
         badfn=None,
     ):
         if opts is None:
             opts = {}
         match = orig(ctx, pats, opts, globbed, default, badfn=badfn)
         return composenormalfilematcher(match, manifest)
 
     with extensions.wrappedfunction(scmutil, b'match', normalfilesmatchfn):
         try:
             result = orig(ui, repo, pats, opts, rename)
         except error.Abort as e:
             if pycompat.bytestr(e) != _(b'no files to copy'):
                 raise e
             else:
                 nonormalfiles = True
             result = 0
 
     # The first rename can cause our current working directory to be removed.
     # In that case there is nothing left to copy/rename so just quit.
     try:
         repo.getcwd()
     except OSError:
         return result
 
     def makestandin(relpath):
         path = pathutil.canonpath(repo.root, repo.getcwd(), relpath)
         return repo.wvfs.join(lfutil.standin(path))
 
     fullpats = scmutil.expandpats(pats)
     dest = fullpats[-1]
 
     if os.path.isdir(dest):
         if not os.path.isdir(makestandin(dest)):
             os.makedirs(makestandin(dest))
 
     try:
         # When we call orig below it creates the standins but we don't add
         # them to the dir state until later so lock during that time.
         wlock = repo.wlock()
 
         manifest = repo[None].manifest()
 
         def overridematch(
             orig,
             ctx,
             pats=(),
             opts=None,
             globbed=False,
             default=b'relpath',
             badfn=None,
         ):
             if opts is None:
                 opts = {}
             newpats = []
             # The patterns were previously mangled to add the standin
             # directory; we need to remove that now
             for pat in pats:
                 if matchmod.patkind(pat) is None and lfutil.shortname in pat:
                     newpats.append(pat.replace(lfutil.shortname, b''))
                 else:
                     newpats.append(pat)
             match = orig(ctx, newpats, opts, globbed, default, badfn=badfn)
             m = copy.copy(match)
             lfile = lambda f: lfutil.standin(f) in manifest
             m._files = [lfutil.standin(f) for f in m._files if lfile(f)]
             m._fileset = set(m._files)
             origmatchfn = m.matchfn
 
             def matchfn(f):
                 lfile = lfutil.splitstandin(f)
                 return (
                     lfile is not None
                     and (f in manifest)
                     and origmatchfn(lfile)
                     or None
                 )
 
             m.matchfn = matchfn
             return m
 
         listpats = []
         for pat in pats:
             if matchmod.patkind(pat) is not None:
                 listpats.append(pat)
             else:
                 listpats.append(makestandin(pat))
 
         copiedfiles = []
 
         def overridecopyfile(orig, src, dest, *args, **kwargs):
             if lfutil.shortname in src and dest.startswith(
                 repo.wjoin(lfutil.shortname)
             ):
                 destlfile = dest.replace(lfutil.shortname, b'')
                 if not opts[b'force'] and os.path.exists(destlfile):
                     raise IOError(
                         b'', _(b'destination largefile already exists')
                     )
             copiedfiles.append((src, dest))
             orig(src, dest, *args, **kwargs)
 
         with extensions.wrappedfunction(util, b'copyfile', overridecopyfile):
             with extensions.wrappedfunction(scmutil, b'match', overridematch):
                 result += orig(ui, repo, listpats, opts, rename)
 
         lfdirstate = lfutil.openlfdirstate(ui, repo)
         for (src, dest) in copiedfiles:
             if lfutil.shortname in src and dest.startswith(
                 repo.wjoin(lfutil.shortname)
             ):
                 srclfile = src.replace(repo.wjoin(lfutil.standin(b'')), b'')
                 destlfile = dest.replace(repo.wjoin(lfutil.standin(b'')), b'')
                 destlfiledir = repo.wvfs.dirname(repo.wjoin(destlfile)) or b'.'
                 if not os.path.isdir(destlfiledir):
                     os.makedirs(destlfiledir)
                 if rename:
                     os.rename(repo.wjoin(srclfile), repo.wjoin(destlfile))
 
                     # The file is gone, but this deletes any empty parent
                     # directories as a side-effect.
                     repo.wvfs.unlinkpath(srclfile, ignoremissing=True)
                     lfdirstate.remove(srclfile)
                 else:
                     util.copyfile(repo.wjoin(srclfile), repo.wjoin(destlfile))
 
                 lfdirstate.add(destlfile)
         lfdirstate.write()
     except error.Abort as e:
         if pycompat.bytestr(e) != _(b'no files to copy'):
             raise e
         else:
             nolfiles = True
     finally:
         wlock.release()
 
     if nolfiles and nonormalfiles:
         raise error.Abort(_(b'no files to copy'))
 
     return result
 
 
 # When the user calls revert, we have to be careful to not revert any
 # changes to other largefiles accidentally. This means we have to keep
 # track of the largefiles that are being reverted so we only pull down
 # the necessary largefiles.
 #
 # Standins are only updated (to match the hash of largefiles) before
 # commits. Update the standins then run the original revert, changing
 # the matcher to hit standins instead of largefiles. Based on the
 # resulting standins update the largefiles.
 @eh.wrapfunction(cmdutil, b'revert')
 def overriderevert(orig, ui, repo, ctx, parents, *pats, **opts):
     # Because we put the standins in a bad state (by updating them)
     # and then return them to a correct state we need to lock to
     # prevent others from changing them in their incorrect state.
     with repo.wlock():
         lfdirstate = lfutil.openlfdirstate(ui, repo)
         s = lfutil.lfdirstatestatus(lfdirstate, repo)
         lfdirstate.write()
         for lfile in s.modified:
             lfutil.updatestandin(repo, lfile, lfutil.standin(lfile))
         for lfile in s.deleted:
             fstandin = lfutil.standin(lfile)
             if repo.wvfs.exists(fstandin):
                 repo.wvfs.unlink(fstandin)
 
         oldstandins = lfutil.getstandinsstate(repo)
 
         def overridematch(
             orig,
             mctx,
             pats=(),
             opts=None,
             globbed=False,
             default=b'relpath',
             badfn=None,
         ):
             if opts is None:
                 opts = {}
             match = orig(mctx, pats, opts, globbed, default, badfn=badfn)
             m = copy.copy(match)
 
             # revert supports recursing into subrepos, and though largefiles
             # currently doesn't work correctly in that case, this match is
             # called, so the lfdirstate above may not be the correct one for
             # this invocation of match.
             lfdirstate = lfutil.openlfdirstate(
                 mctx.repo().ui, mctx.repo(), False
             )
 
             wctx = repo[None]
             matchfiles = []
             for f in m._files:
                 standin = lfutil.standin(f)
                 if standin in ctx or standin in mctx:
                     matchfiles.append(standin)
                 elif standin in wctx or lfdirstate[f] == b'r':
                     continue
                 else:
                     matchfiles.append(f)
             m._files = matchfiles
             m._fileset = set(m._files)
             origmatchfn = m.matchfn
 
             def matchfn(f):
                 lfile = lfutil.splitstandin(f)
                 if lfile is not None:
                     return origmatchfn(lfile) and (f in ctx or f in mctx)
                 return origmatchfn(f)
 
             m.matchfn = matchfn
             return m
 
         with extensions.wrappedfunction(scmutil, b'match', overridematch):
             orig(ui, repo, ctx, parents, *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(r'lfrev', [])
     if opts.get(r'all_largefiles'):
         lfrevs.append(b'pulled()')
     if lfrevs and revspostpull > revsprepull:
         numcached = 0
         repo.firstpulled = revsprepull  # for pulled() revset expression
         try:
             for rev in scmutil.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(r'lfrev', None)
     if lfrevs:
         opargs = kwargs.setdefault(r'opargs', {})
         opargs[b'lfrevs'] = scmutil.revrange(repo, lfrevs)
     return orig(ui, repo, *args, **kwargs)
 
 
 @eh.wrapfunction(exchange, b'pushoperation')
 def exchangepushoperation(orig, *args, **kwargs):
     """Override pushoperation constructor and store lfrevs parameter"""
     lfrevs = kwargs.pop(r'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(r'all_largefiles') and not hg.islocal(d):
         raise error.Abort(
             _(b'--all-largefiles is incompatible with non-local destination %s')
             % d
         )
 
     return orig(ui, source, dest, **opts)
 
 
 @eh.wrapfunction(hg, b'clone')
 def hgclone(orig, ui, opts, *args, **kwargs):
     result = orig(ui, opts, *args, **kwargs)
 
     if result is not None:
         sourcerepo, destrepo = result
         repo = destrepo.local()
 
         # When cloning to a remote repo (like through SSH), no repo is available
         # from the peer.   Therefore the largefiles can't be downloaded and the
         # hgrc can't be updated.
         if not repo:
             return result
 
         # Caching is implicitly limited to 'rev' option, since the dest repo was
         # truncated at that point.  The user may expect a download count with
         # this option, so attempt whether or not this is a largefile repo.
         if opts.get(b'all_largefiles'):
             success, missing = lfcommands.downloadlfiles(ui, repo, None)
 
             if missing != 0:
                 return None
 
     return result
 
 
 @eh.wrapcommand(b'rebase', extension=b'rebase')
 def overriderebase(orig, ui, repo, **opts):
     if not util.safehasattr(repo, b'_largefilesenabled'):
         return orig(ui, repo, **opts)
 
     resuming = opts.get(r'continue')
     repo._lfcommithooks.append(lfutil.automatedcommithook(resuming))
     repo._lfstatuswriters.append(lambda *msg, **opts: None)
     try:
         return orig(ui, repo, **opts)
     finally:
         repo._lfstatuswriters.pop()
         repo._lfcommithooks.pop()
 
 
 @eh.wrapcommand(b'archive')
 def overridearchivecmd(orig, ui, repo, dest, **opts):
     with lfstatus(repo.unfiltered()):
         return orig(ui, repo.unfiltered(), dest, **opts)
 
 
 @eh.wrapfunction(webcommands, b'archive')
 def hgwebarchive(orig, web):
     with lfstatus(web.repo):
         return orig(web)
 
 
 @eh.wrapfunction(archival, b'archive')
 def overridearchive(
     orig,
     repo,
     dest,
     node,
     kind,
     decode=True,
     match=None,
     prefix=b'',
     mtime=None,
     subrepos=None,
 ):
     # For some reason setting repo.lfstatus in hgwebarchive only changes the
     # unfiltered repo's attr, so check that as well.
     if not repo.lfstatus and not repo.unfiltered().lfstatus:
         return orig(
             repo, dest, node, kind, decode, match, prefix, mtime, subrepos
         )
 
     # No need to lock because we are only reading history and
     # largefile caches, neither of which are modified.
     if node is not None:
         lfcommands.cachelfiles(repo.ui, repo, node)
 
     if kind not in archival.archivers:
         raise error.Abort(_(b"unknown archive type '%s'") % kind)
 
     ctx = repo[node]
 
     if kind == b'files':
         if prefix:
             raise error.Abort(_(b'cannot give prefix when archiving to files'))
     else:
         prefix = archival.tidyprefix(dest, kind, prefix)
 
     def write(name, mode, islink, getdata):
         if match and not match(name):
             return
         data = getdata()
         if decode:
             data = repo.wwritedata(name, data)
         archiver.addfile(prefix + name, mode, islink, data)
 
     archiver = archival.archivers[kind](dest, mtime or ctx.date()[0])
 
     if repo.ui.configbool(b"ui", b"archivemeta"):
         write(
             b'.hg_archival.txt',
             0o644,
             False,
             lambda: archival.buildmetadata(ctx),
         )
 
     for f in ctx:
         ff = ctx.flags(f)
         getdata = ctx[f].data
         lfile = lfutil.splitstandin(f)
         if lfile is not None:
             if node is not None:
                 path = lfutil.findfile(repo, getdata().strip())
 
                 if path is None:
                     raise error.Abort(
                         _(
                             b'largefile %s not found in repo store or system cache'
                         )
                         % lfile
                     )
             else:
                 path = lfile
 
             f = lfile
 
             getdata = lambda: util.readfile(path)
         write(f, b'x' in ff and 0o755 or 0o644, b'l' in ff, getdata)
 
     if subrepos:
         for subpath in sorted(ctx.substate):
             sub = ctx.workingsub(subpath)
             submatch = matchmod.subdirmatcher(subpath, match)
             subprefix = prefix + subpath + b'/'
-            with lfstatus(sub._repo):
+
+            # TODO: Only hgsubrepo instances have `_repo`, so figure out how to
+            # infer and possibly set lfstatus in hgsubrepoarchive.  That would
+            # allow only hgsubrepos to set this, instead of the current scheme
+            # where the parent sets this for the child.
+            with (
+                util.safehasattr(sub, '_repo')
+                and lfstatus(sub._repo)
+                or util.nullcontextmanager()
+            ):
                 sub.archive(archiver, subprefix, submatch)
 
     archiver.done()
 
 
 @eh.wrapfunction(subrepo.hgsubrepo, b'archive')
 def hgsubrepoarchive(orig, repo, archiver, prefix, match=None, decode=True):
     lfenabled = util.safehasattr(repo._repo, b'_largefilesenabled')
     if not lfenabled or not repo._repo.lfstatus:
         return orig(repo, archiver, prefix, match, decode)
 
     repo._get(repo._state + (b'hg',))
     rev = repo._state[1]
     ctx = repo._repo[rev]
 
     if ctx.node() is not None:
         lfcommands.cachelfiles(repo.ui, repo._repo, ctx.node())
 
     def write(name, mode, islink, getdata):
         # At this point, the standin has been replaced with the largefile name,
         # so the normal matcher works here without the lfutil variants.
         if match and not match(f):
             return
         data = getdata()
         if decode:
             data = repo._repo.wwritedata(name, data)
 
         archiver.addfile(prefix + name, mode, islink, data)
 
     for f in ctx:
         ff = ctx.flags(f)
         getdata = ctx[f].data
         lfile = lfutil.splitstandin(f)
         if lfile is not None:
             if ctx.node() is not None:
                 path = lfutil.findfile(repo._repo, getdata().strip())
 
                 if path is None:
                     raise error.Abort(
                         _(
                             b'largefile %s not found in repo store or system cache'
                         )
                         % lfile
                     )
             else:
                 path = lfile
 
             f = lfile
 
             getdata = lambda: util.readfile(os.path.join(prefix, path))
 
         write(f, b'x' in ff and 0o755 or 0o644, b'l' in ff, getdata)
 
     for subpath in sorted(ctx.substate):
         sub = ctx.workingsub(subpath)
         submatch = matchmod.subdirmatcher(subpath, match)
         subprefix = prefix + subpath + b'/'
-        with lfstatus(sub._repo):
+        # TODO: Only hgsubrepo instances have `_repo`, so figure out how to
+        # infer and possibly set lfstatus at the top of this function.  That
+        # would allow only hgsubrepos to set this, instead of the current scheme
+        # where the parent sets this for the child.
+        with (
+            util.safehasattr(sub, '_repo')
+            and lfstatus(sub._repo)
+            or util.nullcontextmanager()
+        ):
             sub.archive(archiver, subprefix, submatch, decode)
 
 
 # If a largefile is modified, the change is not reflected in its
 # standin until a commit. cmdutil.bailifchanged() raises an exception
 # if the repo has uncommitted changes. Wrap it to also check if
 # largefiles were changed. This is used by bisect, backout and fetch.
 @eh.wrapfunction(cmdutil, b'bailifchanged')
 def overridebailifchanged(orig, repo, *args, **kwargs):
     orig(repo, *args, **kwargs)
     with lfstatus(repo):
         s = repo.status()
     if s.modified or s.added or s.removed or s.deleted:
         raise error.Abort(_(b'uncommitted changes'))
 
 
 @eh.wrapfunction(cmdutil, b'postcommitstatus')
 def postcommitstatus(orig, repo, *args, **kwargs):
     with lfstatus(repo):
         return orig(repo, *args, **kwargs)
 
 
 @eh.wrapfunction(cmdutil, b'forget')
 def cmdutilforget(
     orig, ui, repo, match, prefix, uipathfn, explicitonly, dryrun, interactive
 ):
     normalmatcher = composenormalfilematcher(match, repo[None].manifest())
     bad, forgot = orig(
         ui,
         repo,
         normalmatcher,
         prefix,
         uipathfn,
         explicitonly,
         dryrun,
         interactive,
     )
     m = composelargefilematcher(match, repo[None].manifest())
 
     with lfstatus(repo):
         s = repo.status(match=m, clean=True)
     manifest = repo[None].manifest()
     forget = sorted(s.modified + s.added + s.deleted + s.clean)
     forget = [f for f in forget if lfutil.standin(f) in manifest]
 
     for f in forget:
         fstandin = lfutil.standin(f)
         if fstandin not in repo.dirstate and not repo.wvfs.isdir(fstandin):
             ui.warn(
                 _(b'not removing %s: file is already untracked\n') % uipathfn(f)
             )
             bad.append(f)
 
     for f in forget:
         if ui.verbose or not m.exact(f):
             ui.status(_(b'removing %s\n') % uipathfn(f))
 
     # Need to lock because standin files are deleted then removed from the
     # repository and we could race in-between.
     with repo.wlock():
         lfdirstate = lfutil.openlfdirstate(ui, repo)
         for f in forget:
             if lfdirstate[f] == b'a':
                 lfdirstate.drop(f)
             else:
                 lfdirstate.remove(f)
         lfdirstate.write()
         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] = []
                 toupload[fn].append(lfhash)
                 lfhashes.add(lfhash)
 
             def showhashes(fn):
                 for lfhash in sorted(toupload[fn]):
                     ui.debug(b'    %s\n' % lfhash)
 
         else:
             toupload = set()
 
             def addfunc(fn, lfhash):
                 toupload.add(fn)
                 lfhashes.add(lfhash)
 
             def showhashes(fn):
                 pass
 
         _getoutgoings(repo, other, missing, addfunc)
 
         if not toupload:
             ui.status(_(b'largefiles: no files to upload\n'))
         else:
             ui.status(
                 _(b'largefiles to upload (%d entities):\n') % (len(lfhashes))
             )
             for file in sorted(toupload):
                 ui.status(lfutil.splitstandin(file) + b'\n')
                 showhashes(file)
             ui.status(b'\n')
 
 
 @eh.wrapcommand(
     b'outgoing', opts=[(b'', b'large', None, _(b'display outgoing largefiles'))]
 )
 def _outgoingcmd(orig, *args, **kwargs):
     # Nothing to do here other than add the extra help option- the hook above
     # processes it.
     return orig(*args, **kwargs)
 
 
 def summaryremotehook(ui, repo, opts, changes):
     largeopt = opts.get(b'large', False)
     if changes is None:
         if largeopt:
             return (False, True)  # only outgoing check is needed
         else:
             return (False, False)
     elif largeopt:
         url, branch, peer, outgoing = changes[1]
         if peer is None:
             # i18n: column positioning for "hg summary"
             ui.status(_(b'largefiles: (no remote repo)\n'))
             return
 
         toupload = set()
         lfhashes = set()
 
         def addfunc(fn, lfhash):
             toupload.add(fn)
             lfhashes.add(lfhash)
 
         _getoutgoings(repo, peer, outgoing.missing, addfunc)
 
         if not toupload:
             # i18n: column positioning for "hg summary"
             ui.status(_(b'largefiles: (no files to upload)\n'))
         else:
             # i18n: column positioning for "hg summary"
             ui.status(
                 _(b'largefiles: %d entities for %d files to upload\n')
                 % (len(lfhashes), len(toupload))
             )
 
 
 @eh.wrapcommand(
     b'summary', opts=[(b'', b'large', None, _(b'display outgoing largefiles'))]
 )
 def overridesummary(orig, ui, repo, *pats, **opts):
     with lfstatus(repo):
         orig(ui, repo, *pats, **opts)
 
 
 @eh.wrapfunction(scmutil, b'addremove')
 def scmutiladdremove(orig, repo, matcher, prefix, uipathfn, opts=None):
     if opts is None:
         opts = {}
     if not lfutil.islfilesrepo(repo):
         return orig(repo, matcher, prefix, uipathfn, opts)
     # Get the list of missing largefiles so we can remove them
     lfdirstate = lfutil.openlfdirstate(repo.ui, repo)
     unsure, s = 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)
 
         # 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)
 
 
 # Calling purge with --all will cause the largefiles to be deleted.
 # Override repo.status to prevent this from happening.
 @eh.wrapcommand(b'purge', extension=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 lfdirstate[f] == b'?']
         ignored = [f for f in r.ignored if lfdirstate[f] == b'?']
         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 = set(
             f
             for f in repo.dirstate
             if lfutil.isstandin(f) and repo.dirstate[f] != b'r'
         )
         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[f] == b'r':
                     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)
 
         lfdirstate = lfutil.openlfdirstate(ui, repo)
         orphans = set(lfdirstate)
         lfiles = lfutil.listlfiles(repo)
         for file in lfiles:
             lfutil.synclfdirstate(repo, lfdirstate, file, True)
             orphans.discard(file)
         for lfile in orphans:
             lfdirstate.drop(lfile)
         lfdirstate.write()
     return result
 
 
 @eh.wrapcommand(b'transplant', extension=b'transplant')
 def overridetransplant(orig, ui, repo, *revs, **opts):
     resuming = opts.get(r'continue')
     repo._lfcommithooks.append(lfutil.automatedcommithook(resuming))
     repo._lfstatuswriters.append(lambda *msg, **opts: None)
     try:
         result = orig(ui, repo, *revs, **opts)
     finally:
         repo._lfstatuswriters.pop()
         repo._lfcommithooks.pop()
     return result
 
 
 @eh.wrapcommand(b'cat')
 def overridecat(orig, ui, repo, file1, *pats, **opts):
     opts = pycompat.byteskwargs(opts)
     ctx = scmutil.revsingle(repo, opts.get(b'rev'))
     err = 1
     notbad = set()
     m = scmutil.match(ctx, (file1,) + pats, opts)
     origmatchfn = m.matchfn
 
     def lfmatchfn(f):
         if origmatchfn(f):
             return True
         lf = lfutil.splitstandin(f)
         if lf is None:
             return False
         notbad.add(lf)
         return origmatchfn(lf)
 
     m.matchfn = lfmatchfn
     origbadfn = m.bad
 
     def lfbadfn(f, msg):
         if not f in notbad:
             origbadfn(f, msg)
 
     m.bad = lfbadfn
 
     origvisitdirfn = m.visitdir
 
     def lfvisitdirfn(dir):
         if dir == lfutil.shortname:
             return True
         ret = origvisitdirfn(dir)
         if ret:
             return ret
         lf = lfutil.splitstandin(dir)
         if lf is None:
             return False
         return origvisitdirfn(lf)
 
     m.visitdir = lfvisitdirfn
 
     for f in ctx.walk(m):
         with cmdutil.makefileobj(ctx, opts.get(b'output'), pathname=f) as fp:
             lf = lfutil.splitstandin(f)
             if lf is None or origmatchfn(f):
                 # duplicating unreachable code from commands.cat
                 data = ctx[f].data()
                 if opts.get(b'decode'):
                     data = repo.wwritedata(f, data)
                 fp.write(data)
             else:
                 hash = lfutil.readasstandin(ctx[f])
                 if not lfutil.inusercache(repo.ui, hash):
                     store = storefactory.openstore(repo)
                     success, missing = store.get([(lf, hash)])
                     if len(success) != 1:
                         raise error.Abort(
                             _(
                                 b'largefile %s is not in cache and could not be '
                                 b'downloaded'
                             )
                             % lf
                         )
                 path = lfutil.usercachepath(repo.ui, hash)
                 with open(path, b"rb") as fpin:
                     for chunk in util.filechunkiter(fpin):
                         fp.write(chunk)
         err = 0
     return err
 
 
 @eh.wrapfunction(merge, b'update')
 def mergeupdate(orig, repo, node, branchmerge, force, *args, **kwargs):
     matcher = kwargs.get(r'matcher', None)
     # note if this is a partial update
     partial = matcher and not matcher.always()
     with repo.wlock():
         # branch |       |         |
         #  merge | force | partial | action
         # -------+-------+---------+--------------
         #    x   |   x   |    x    | linear-merge
         #    o   |   x   |    x    | branch-merge
         #    x   |   o   |    x    | overwrite (as clean update)
         #    o   |   o   |    x    | force-branch-merge (*1)
         #    x   |   x   |    o    |   (*)
         #    o   |   x   |    o    |   (*)
         #    x   |   o   |    o    | overwrite (as revert)
         #    o   |   o   |    o    |   (*)
         #
         # (*) don't care
         # (*1) deprecated, but used internally (e.g: "rebase --collapse")
 
         lfdirstate = lfutil.openlfdirstate(repo.ui, repo)
         unsure, s = lfdirstate.status(
             matchmod.always(),
             subrepos=[],
             ignored=False,
             clean=True,
             unknown=False,
         )
         oldclean = set(s.clean)
         pctx = repo[b'.']
         dctx = repo[node]
         for lfile in unsure + s.modified:
             lfileabs = repo.wvfs.join(lfile)
             if not repo.wvfs.exists(lfileabs):
                 continue
             lfhash = lfutil.hashfile(lfileabs)
             standin = lfutil.standin(lfile)
             lfutil.writestandin(
                 repo, standin, lfhash, lfutil.getexecutable(lfileabs)
             )
             if standin in pctx and lfhash == lfutil.readasstandin(
                 pctx[standin]
             ):
                 oldclean.add(lfile)
         for lfile in s.added:
             fstandin = lfutil.standin(lfile)
             if fstandin not in dctx:
                 # in this case, content of standin file is meaningless
                 # (in dctx, lfile is unknown, or normal file)
                 continue
             lfutil.updatestandin(repo, lfile, fstandin)
         # mark all clean largefiles as dirty, just in case the update gets
         # interrupted before largefiles and lfdirstate are synchronized
         for lfile in oldclean:
             lfdirstate.normallookup(lfile)
         lfdirstate.write()
 
         oldstandins = lfutil.getstandinsstate(repo)
         # Make sure the merge runs on disk, not in-memory. largefiles is not a
         # good candidate for in-memory merge (large files, custom dirstate,
         # matcher usage).
         kwargs[r'wc'] = repo[None]
         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.normal(lfile)
         lfdirstate.write()
 
         if branchmerge or force or partial:
             filelist.extend(s.deleted + s.removed)
 
         lfcommands.updatelfiles(
             repo.ui, repo, filelist=filelist, normallookup=partial
         )
 
         return result
 
 
 @eh.wrapfunction(scmutil, b'marktouched')
 def scmutilmarktouched(orig, repo, files, *args, **kwargs):
     result = orig(repo, files, *args, **kwargs)
 
     filelist = []
     for f in files:
         lf = lfutil.splitstandin(f)
         if lf is not None:
             filelist.append(lf)
     if filelist:
         lfcommands.updatelfiles(
             repo.ui,
             repo,
             filelist=filelist,
             printmessage=False,
             normallookup=True,
         )
 
     return result
 
 
 @eh.wrapfunction(upgrade, b'preservedrequirements')
 @eh.wrapfunction(upgrade, b'supporteddestrequirements')
 def upgraderequirements(orig, repo):
     reqs = orig(repo)
     if b'largefiles' in repo.requirements:
         reqs.add(b'largefiles')
     return reqs
 
 
 _lfscheme = b'largefile://'
 
 
 @eh.wrapfunction(urlmod, b'open')
 def openlargefile(orig, ui, url_, data=None):
     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)

tests/test-subrepo-git.t

@@ -1,1262 +1,1271 b''
 #require git
 
 make git commits repeatable
 
   $ cat >> $HGRCPATH <<EOF
   > [defaults]
   > commit = -d "0 0"
   > EOF
 
   $ echo "[core]" >> $HOME/.gitconfig
   $ echo "autocrlf = false" >> $HOME/.gitconfig
   $ GIT_AUTHOR_NAME='test'; export GIT_AUTHOR_NAME
   $ GIT_AUTHOR_EMAIL='[e-mail]'; export GIT_AUTHOR_EMAIL
   $ GIT_AUTHOR_DATE='1234567891 +0000'; export GIT_AUTHOR_DATE
   $ GIT_COMMITTER_NAME="$GIT_AUTHOR_NAME"; export GIT_COMMITTER_NAME
   $ GIT_COMMITTER_EMAIL="$GIT_AUTHOR_EMAIL"; export GIT_COMMITTER_EMAIL
   $ GIT_COMMITTER_DATE="$GIT_AUTHOR_DATE"; export GIT_COMMITTER_DATE
   $ GIT_CONFIG_NOSYSTEM=1; export GIT_CONFIG_NOSYSTEM
 
 root hg repo
 
   $ hg init t
   $ cd t
   $ echo a > a
   $ hg add a
   $ hg commit -m a
   $ cd ..
 
 new external git repo
 
   $ mkdir gitroot
   $ cd gitroot
   $ git init -q
   $ echo g > g
   $ git add g
   $ git commit -q -m g
 
 add subrepo clone
 
   $ cd ../t
   $ echo 's = [git]../gitroot' > .hgsub
   $ git clone -q ../gitroot s
   $ hg add .hgsub
 
 git subrepo is disabled by default
 
   $ hg commit -m 'new git subrepo'
   abort: git subrepos not allowed
   (see 'hg help config.subrepos' for details)
   [255]
 
 so enable it
 
   $ cat >> $HGRCPATH <<EOF
   > [subrepos]
   > git:allowed = true
   > EOF
 
   $ hg commit -m 'new git subrepo'
 
   $ hg debugsub
   path s
    source   ../gitroot
    revision da5f5b1d8ffcf62fb8327bcd3c89a4367a6018e7
 
 record a new commit from upstream from a different branch
 
   $ cd ../gitroot
   $ git checkout -q -b testing
   $ echo gg >> g
   $ git commit -q -a -m gg
 
   $ cd ../t/s
   $ git pull -q >/dev/null 2>/dev/null
   $ git checkout -q -b testing origin/testing >/dev/null
 
   $ cd ..
   $ hg status --subrepos
   M s/g
   $ hg commit -m 'update git subrepo'
   $ hg debugsub
   path s
    source   ../gitroot
    revision 126f2a14290cd5ce061fdedc430170e8d39e1c5a
 
 make $GITROOT pushable, by replacing it with a clone with nothing checked out
 
   $ cd ..
   $ git clone gitroot gitrootbare --bare -q
   $ rm -rf gitroot
   $ mv gitrootbare gitroot
 
 clone root
 
   $ cd t
   $ hg clone . ../tc 2> /dev/null
   updating to branch default
   cloning subrepo s from $TESTTMP/gitroot
   3 files updated, 0 files merged, 0 files removed, 0 files unresolved
   $ cd ../tc
   $ hg debugsub
   path s
    source   ../gitroot
    revision 126f2a14290cd5ce061fdedc430170e8d39e1c5a
   $ cd ..
 
 clone with subrepo disabled (update should fail)
 
   $ hg clone t -U tc2 --config subrepos.allowed=false
   $ hg update -R tc2 --config subrepos.allowed=false
   abort: subrepos not enabled
   (see 'hg help config.subrepos' for details)
   [255]
   $ ls tc2
   a
 
   $ hg clone t tc3 --config subrepos.allowed=false
   updating to branch default
   abort: subrepos not enabled
   (see 'hg help config.subrepos' for details)
   [255]
   $ ls tc3
   a
 
 update to previous substate
 
   $ cd tc
   $ hg update 1 -q
   $ cat s/g
   g
   $ hg debugsub
   path s
    source   ../gitroot
    revision da5f5b1d8ffcf62fb8327bcd3c89a4367a6018e7
 
 clone root, make local change
 
   $ cd ../t
   $ hg clone . ../ta 2> /dev/null
   updating to branch default
   cloning subrepo s from $TESTTMP/gitroot
   3 files updated, 0 files merged, 0 files removed, 0 files unresolved
 
   $ cd ../ta
   $ echo ggg >> s/g
   $ hg status --subrepos
   M s/g
   $ hg diff --subrepos
   diff --git a/s/g b/s/g
   index 089258f..85341ee 100644
   --- a/s/g
   +++ b/s/g
   @@ -1,2 +1,3 @@
    g
    gg
   +ggg
   $ hg commit --subrepos -m ggg
   committing subrepository s
   $ hg debugsub
   path s
    source   ../gitroot
    revision 79695940086840c99328513acbe35f90fcd55e57
 
 clone root separately, make different local change
 
   $ cd ../t
   $ hg clone . ../tb 2> /dev/null
   updating to branch default
   cloning subrepo s from $TESTTMP/gitroot
   3 files updated, 0 files merged, 0 files removed, 0 files unresolved
 
   $ cd ../tb/s
   $ hg status --subrepos
   $ echo f > f
   $ hg status --subrepos
   ? s/f
   $ hg add .
   adding f
   $ git add f
   $ cd ..
 
   $ hg status --subrepos
   A s/f
   $ hg commit --subrepos -m f
   committing subrepository s
   $ hg debugsub
   path s
    source   ../gitroot
    revision aa84837ccfbdfedcdcdeeedc309d73e6eb069edc
 
 user b push changes
 
   $ hg push 2>/dev/null
   pushing to $TESTTMP/t
   pushing branch testing of subrepository "s"
   searching for changes
   adding changesets
   adding manifests
   adding file changes
   added 1 changesets with 1 changes to 1 files
 
 user a pulls, merges, commits
 
   $ cd ../ta
   $ hg pull
   pulling from $TESTTMP/t
   searching for changes
   adding changesets
   adding manifests
   adding file changes
   added 1 changesets with 1 changes to 1 files (+1 heads)
   new changesets 089416c11d73
   (run 'hg heads' to see heads, 'hg merge' to merge)
   $ hg merge 2>/dev/null
    subrepository s diverged (local revision: 7969594, remote revision: aa84837)
   you can (m)erge, keep (l)ocal [working copy] or keep (r)emote [merge rev].
   what do you want to do? m
   pulling subrepo s from $TESTTMP/gitroot
   0 files updated, 0 files merged, 0 files removed, 0 files unresolved
   (branch merge, don't forget to commit)
   $ hg st --subrepos s
   A s/f
   $ cat s/f
   f
   $ cat s/g
   g
   gg
   ggg
   $ hg commit --subrepos -m 'merge'
   committing subrepository s
   $ hg status --subrepos --rev 1:5
   M .hgsubstate
   M s/g
   A s/f
   $ hg debugsub
   path s
    source   ../gitroot
    revision f47b465e1bce645dbf37232a00574aa1546ca8d3
   $ hg push 2>/dev/null
   pushing to $TESTTMP/t
   pushing branch testing of subrepository "s"
   searching for changes
   adding changesets
   adding manifests
   adding file changes
   added 2 changesets with 2 changes to 1 files
 
 make upstream git changes
 
   $ cd ..
   $ git clone -q gitroot gitclone
   $ cd gitclone
   $ echo ff >> f
   $ git commit -q -a -m ff
   $ echo fff >> f
   $ git commit -q -a -m fff
   $ git push origin testing 2>/dev/null
 
 make and push changes to hg without updating the subrepo
 
   $ cd ../t
   $ hg clone . ../td 2>&1 | egrep -v '^Cloning into|^done\.'
   updating to branch default
   cloning subrepo s from $TESTTMP/gitroot
   checking out detached HEAD in subrepository "s"
   check out a git branch if you intend to make changes
   3 files updated, 0 files merged, 0 files removed, 0 files unresolved
   $ cd ../td
   $ echo aa >> a
   $ hg commit -m aa
   $ hg push
   pushing to $TESTTMP/t
   searching for changes
   adding changesets
   adding manifests
   adding file changes
   added 1 changesets with 1 changes to 1 files
 
 sync to upstream git, distribute changes
 
   $ cd ../ta
   $ hg pull -u -q
   $ cd s
   $ git pull -q >/dev/null 2>/dev/null
   $ cd ..
   $ hg commit -m 'git upstream sync'
   $ hg debugsub
   path s
    source   ../gitroot
    revision 32a343883b74769118bb1d3b4b1fbf9156f4dddc
   $ hg push -q
 
   $ cd ../tb
   $ hg pull -q
   $ hg update 2>/dev/null
   pulling subrepo s from $TESTTMP/gitroot
   2 files updated, 0 files merged, 0 files removed, 0 files unresolved
   $ hg debugsub
   path s
    source   ../gitroot
    revision 32a343883b74769118bb1d3b4b1fbf9156f4dddc
 
 create a new git branch
 
   $ cd s
   $ git checkout -b b2
   Switched to a new branch 'b2'
   $ echo a>a
   $ git add a
   $ git commit -qm 'add a'
   $ cd ..
   $ hg commit -m 'add branch in s'
 
 pulling new git branch should not create tracking branch named 'origin/b2'
 (issue3870)
   $ cd ../td/s
   $ git remote set-url origin $TESTTMP/tb/s
   $ git branch --no-track oldtesting
   $ cd ..
   $ hg pull -q ../tb
   $ hg up
   From $TESTTMP/tb/s
    * [new branch]      b2         -> origin/b2
   Previous HEAD position was f47b465* merge (glob)
   Switched to a new branch 'b2'
   pulling subrepo s from $TESTTMP/tb/s
   1 files updated, 0 files merged, 0 files removed, 0 files unresolved
 
 update to a revision without the subrepo, keeping the local git repository
 
   $ cd ../t
   $ hg up 0
   0 files updated, 0 files merged, 2 files removed, 0 files unresolved
   $ ls -a s
   .
   ..
   .git
 
   $ hg up 2
   2 files updated, 0 files merged, 0 files removed, 0 files unresolved
   $ ls -a s
   .
   ..
   .git
   g
 
 archive subrepos
 
   $ cd ../tc
   $ hg pull -q
   $ hg archive --subrepos -r 5 ../archive 2>/dev/null
   pulling subrepo s from $TESTTMP/gitroot
   $ cd ../archive
   $ cat s/f
   f
   $ cat s/g
   g
   gg
   ggg
 
   $ hg -R ../tc archive --subrepo -r 5 -X ../tc/**f ../archive_x 2>/dev/null
   $ find ../archive_x | sort | grep -v pax_global_header
   ../archive_x
   ../archive_x/.hg_archival.txt
   ../archive_x/.hgsub
   ../archive_x/.hgsubstate
   ../archive_x/a
   ../archive_x/s
   ../archive_x/s/g
 
   $ hg -R ../tc archive -S ../archive.tgz --prefix '.' 2>/dev/null
   $ tar -tzf ../archive.tgz | sort | grep -v pax_global_header
   .hg_archival.txt
   .hgsub
   .hgsubstate
   a
   s/g
 
+  $ hg -R ../tc archive -S ../lf_archive.tgz --prefix '.' \
+  >       --config extensions.largefiles= 2>/dev/null
+  $ tar -tzf ../lf_archive.tgz | sort | grep -v pax_global_header
+  .hg_archival.txt
+  .hgsub
+  .hgsubstate
+  a
+  s/g
+
 create nested repo
 
   $ cd ..
   $ hg init outer
   $ cd outer
   $ echo b>b
   $ hg add b
   $ hg commit -m b
 
   $ hg clone ../t inner 2> /dev/null
   updating to branch default
   cloning subrepo s from $TESTTMP/gitroot
   3 files updated, 0 files merged, 0 files removed, 0 files unresolved
   $ echo inner = inner > .hgsub
   $ hg add .hgsub
   $ hg commit -m 'nested sub'
 
 nested commit
 
   $ echo ffff >> inner/s/f
   $ hg status --subrepos
   M inner/s/f
   $ hg commit --subrepos -m nested
   committing subrepository inner
   committing subrepository inner/s
 
 nested archive
 
   $ hg archive --subrepos ../narchive
   $ ls ../narchive/inner/s | grep -v pax_global_header
   f
   g
 
 relative source expansion
 
   $ cd ..
   $ mkdir d
   $ hg clone t d/t 2> /dev/null
   updating to branch default
   cloning subrepo s from $TESTTMP/gitroot
   3 files updated, 0 files merged, 0 files removed, 0 files unresolved
 
 Don't crash if the subrepo is missing
 
   $ hg clone t missing -q
   $ cd missing
   $ rm -rf s
   $ hg status -S
   $ hg sum | grep commit
   commit: 1 subrepos
   $ hg push -q
   abort: subrepo s is missing (in subrepository "s")
   [255]
   $ hg commit --subrepos -qm missing
   abort: subrepo s is missing (in subrepository "s")
   [255]
 
 #if symlink
 Don't crash if subrepo is a broken symlink
   $ ln -s broken s
   $ hg status -S
   abort: subrepo 's' traverses symbolic link
   [255]
   $ hg push -q
   abort: subrepo 's' traverses symbolic link
   [255]
   $ hg commit --subrepos -qm missing
   abort: subrepo 's' traverses symbolic link
   [255]
   $ rm s
 #endif
 
   $ hg update -C 2> /dev/null
   cloning subrepo s from $TESTTMP/gitroot
   1 files updated, 0 files merged, 0 files removed, 0 files unresolved
   $ hg sum | grep commit
   commit: (clean)
 
 Don't crash if the .hgsubstate entry is missing
 
   $ hg update 1 -q
   $ hg rm .hgsubstate
   $ hg commit .hgsubstate -m 'no substate'
   nothing changed
   [1]
   $ hg tag -l nosubstate
   $ hg manifest
   .hgsub
   .hgsubstate
   a
 
   $ hg status -S
   R .hgsubstate
   $ hg sum | grep commit
   commit: 1 removed, 1 subrepos (new branch head)
 
   $ hg commit -m 'restore substate'
   nothing changed
   [1]
   $ hg manifest
   .hgsub
   .hgsubstate
   a
   $ hg sum | grep commit
   commit: 1 removed, 1 subrepos (new branch head)
 
   $ hg update -qC nosubstate
   $ ls s
   g
 
 issue3109: false positives in git diff-index
 
   $ hg update -q
   $ touch -t 200001010000 s/g
   $ hg status --subrepos
   $ touch -t 200001010000 s/g
   $ hg sum | grep commit
   commit: (clean)
 
 Check hg update --clean
   $ cd $TESTTMP/ta
   $ echo  > s/g
   $ cd s
   $ echo c1 > f1
   $ echo c1 > f2
   $ git add f1
   $ cd ..
   $ hg status -S
   M s/g
   A s/f1
   ? s/f2
   $ ls s
   f
   f1
   f2
   g
   $ hg update --clean
   1 files updated, 0 files merged, 0 files removed, 0 files unresolved
   $ hg status -S
   ? s/f1
   ? s/f2
   $ ls s
   f
   f1
   f2
   g
 
 Sticky subrepositories, no changes
   $ cd $TESTTMP/ta
   $ hg id -n
   7
   $ cd s
   $ git rev-parse HEAD
   32a343883b74769118bb1d3b4b1fbf9156f4dddc
   $ cd ..
   $ hg update 1 > /dev/null 2>&1
   $ hg id -n
   1
   $ cd s
   $ git rev-parse HEAD
   da5f5b1d8ffcf62fb8327bcd3c89a4367a6018e7
   $ cd ..
 
 Sticky subrepositories, file changes
   $ touch s/f1
   $ cd s
   $ git add f1
   $ cd ..
   $ hg id -n
   1+
   $ cd s
   $ git rev-parse HEAD
   da5f5b1d8ffcf62fb8327bcd3c89a4367a6018e7
   $ cd ..
   $ hg update 4
    subrepository s diverged (local revision: da5f5b1, remote revision: aa84837)
   you can (m)erge, keep (l)ocal [working copy] or keep (r)emote [destination].
   what do you want to do? m
    subrepository sources for s differ
   you can use (l)ocal source (da5f5b1) or (r)emote source (aa84837).
   what do you want to do? l
   0 files updated, 0 files merged, 0 files removed, 0 files unresolved
   $ hg id -n
   4+
   $ cd s
   $ git rev-parse HEAD
   da5f5b1d8ffcf62fb8327bcd3c89a4367a6018e7
   $ cd ..
   $ hg update --clean tip > /dev/null 2>&1
 
 Sticky subrepository, revision updates
   $ hg id -n
   7
   $ cd s
   $ git rev-parse HEAD
   32a343883b74769118bb1d3b4b1fbf9156f4dddc
   $ cd ..
   $ cd s
   $ git checkout aa84837ccfbdfedcdcdeeedc309d73e6eb069edc
   Previous HEAD position was 32a3438* fff (glob)
   HEAD is now at aa84837* f (glob)
   $ cd ..
   $ hg update 1
    subrepository s diverged (local revision: 32a3438, remote revision: da5f5b1)
   you can (m)erge, keep (l)ocal [working copy] or keep (r)emote [destination].
   what do you want to do? m
    subrepository sources for s differ (in checked out version)
   you can use (l)ocal source (32a3438) or (r)emote source (da5f5b1).
   what do you want to do? l
   1 files updated, 0 files merged, 0 files removed, 0 files unresolved
   $ hg id -n
   1+
   $ cd s
   $ git rev-parse HEAD
   aa84837ccfbdfedcdcdeeedc309d73e6eb069edc
   $ cd ..
 
 Sticky subrepository, file changes and revision updates
   $ touch s/f1
   $ cd s
   $ git add f1
   $ git rev-parse HEAD
   aa84837ccfbdfedcdcdeeedc309d73e6eb069edc
   $ cd ..
   $ hg id -n
   1+
   $ hg update 7
    subrepository s diverged (local revision: 32a3438, remote revision: 32a3438)
   you can (m)erge, keep (l)ocal [working copy] or keep (r)emote [destination].
   what do you want to do? m
    subrepository sources for s differ
   you can use (l)ocal source (32a3438) or (r)emote source (32a3438).
   what do you want to do? l
   1 files updated, 0 files merged, 0 files removed, 0 files unresolved
   $ hg id -n
   7+
   $ cd s
   $ git rev-parse HEAD
   aa84837ccfbdfedcdcdeeedc309d73e6eb069edc
   $ cd ..
 
 Sticky repository, update --clean
   $ hg update --clean tip 2>/dev/null
   1 files updated, 0 files merged, 0 files removed, 0 files unresolved
   $ hg id -n
   7
   $ cd s
   $ git rev-parse HEAD
   32a343883b74769118bb1d3b4b1fbf9156f4dddc
   $ cd ..
 
 Test subrepo already at intended revision:
   $ cd s
   $ git checkout 32a343883b74769118bb1d3b4b1fbf9156f4dddc
   HEAD is now at 32a3438* fff (glob)
   $ cd ..
   $ hg update 1
   Previous HEAD position was 32a3438* fff (glob)
   HEAD is now at da5f5b1* g (glob)
   2 files updated, 0 files merged, 0 files removed, 0 files unresolved
   $ hg id -n
   1
   $ cd s
   $ git rev-parse HEAD
   da5f5b1d8ffcf62fb8327bcd3c89a4367a6018e7
   $ cd ..
 
 Test forgetting files, not implemented in git subrepo, used to
 traceback
 #if no-windows
   $ hg forget 'notafile*'
   notafile*: $ENOENT$
   [1]
 #else
 error: The filename, directory name, or volume label syntax is incorrect
   $ hg forget 'notafile'
   notafile: * (glob)
   [1]
 #endif
 
   $ cd ..
 
 Test sanitizing ".hg/hgrc" in subrepo
 
   $ cd t
   $ hg tip -q
   7:af6d2edbb0d3
   $ hg update -q -C af6d2edbb0d3
   $ cd s
   $ git checkout -q -b sanitize-test
   $ mkdir .hg
   $ echo '.hg/hgrc in git repo' > .hg/hgrc
   $ mkdir -p sub/.hg
   $ echo 'sub/.hg/hgrc in git repo' > sub/.hg/hgrc
   $ git add .hg sub
   $ git commit -qm 'add .hg/hgrc to be sanitized at hg update'
   $ git push -q origin sanitize-test
   $ cd ..
   $ grep ' s$' .hgsubstate
   32a343883b74769118bb1d3b4b1fbf9156f4dddc s
   $ hg commit -qm 'commit with git revision including .hg/hgrc'
   $ hg parents -q
   8:3473d20bddcf
   $ grep ' s$' .hgsubstate
   c4069473b459cf27fd4d7c2f50c4346b4e936599 s
   $ cd ..
 
   $ hg -R tc pull -q
   $ hg -R tc update -q -C 3473d20bddcf 2>&1 | sort
   warning: removing potentially hostile 'hgrc' in '$TESTTMP/tc/s/.hg'
   warning: removing potentially hostile 'hgrc' in '$TESTTMP/tc/s/sub/.hg'
   $ cd tc
   $ hg parents -q
   8:3473d20bddcf
   $ grep ' s$' .hgsubstate
   c4069473b459cf27fd4d7c2f50c4346b4e936599 s
   $ test -f s/.hg/hgrc
   [1]
   $ test -f s/sub/.hg/hgrc
   [1]
   $ cd ..
 
 additional test for "git merge --ff" route:
 
   $ cd t
   $ hg tip -q
   8:3473d20bddcf
   $ hg update -q -C af6d2edbb0d3
   $ cd s
   $ git checkout -q testing
   $ mkdir .hg
   $ echo '.hg/hgrc in git repo' > .hg/hgrc
   $ mkdir -p sub/.hg
   $ echo 'sub/.hg/hgrc in git repo' > sub/.hg/hgrc
   $ git add .hg sub
   $ git commit -qm 'add .hg/hgrc to be sanitized at hg update (git merge --ff)'
   $ git push -q origin testing
   $ cd ..
   $ grep ' s$' .hgsubstate
   32a343883b74769118bb1d3b4b1fbf9156f4dddc s
   $ hg commit -qm 'commit with git revision including .hg/hgrc'
   $ hg parents -q
   9:ed23f7fe024e
   $ grep ' s$' .hgsubstate
   f262643c1077219fbd3858d54e78ef050ef84fbf s
   $ cd ..
 
   $ cd tc
   $ hg update -q -C af6d2edbb0d3
   $ test -f s/.hg/hgrc
   [1]
   $ test -f s/sub/.hg/hgrc
   [1]
   $ cd ..
   $ hg -R tc pull -q
   $ hg -R tc update -q -C ed23f7fe024e 2>&1 | sort
   warning: removing potentially hostile 'hgrc' in '$TESTTMP/tc/s/.hg'
   warning: removing potentially hostile 'hgrc' in '$TESTTMP/tc/s/sub/.hg'
   $ cd tc
   $ hg parents -q
   9:ed23f7fe024e
   $ grep ' s$' .hgsubstate
   f262643c1077219fbd3858d54e78ef050ef84fbf s
   $ test -f s/.hg/hgrc
   [1]
   $ test -f s/sub/.hg/hgrc
   [1]
 
 Test that sanitizing is omitted in meta data area:
 
   $ mkdir s/.git/.hg
   $ echo '.hg/hgrc in git metadata area' > s/.git/.hg/hgrc
   $ hg update -q -C af6d2edbb0d3
   checking out detached HEAD in subrepository "s"
   check out a git branch if you intend to make changes
 
 check differences made by most recent change
   $ cd s
   $ cat > foobar << EOF
   > woopwoop
   > 
   > foo
   > bar
   > EOF
   $ git add foobar
   $ cd ..
 
   $ hg diff --subrepos
   diff --git a/s/foobar b/s/foobar
   new file mode 100644
   index 0000000..8a5a5e2
   --- /dev/null
   +++ b/s/foobar
   @@ -0,0 +1,4 @@
   +woopwoop
   +
   +foo
   +bar
 
   $ hg commit --subrepos -m "Added foobar"
   committing subrepository s
   created new head
 
   $ hg diff -c . --subrepos --nodates
   diff -r af6d2edbb0d3 -r 255ee8cf690e .hgsubstate
   --- a/.hgsubstate
   +++ b/.hgsubstate
   @@ -1,1 +1,1 @@
   -32a343883b74769118bb1d3b4b1fbf9156f4dddc s
   +fd4dbf828a5b2fcd36b2bcf21ea773820970d129 s
   diff --git a/s/foobar b/s/foobar
   new file mode 100644
   index 0000000..8a5a5e2
   --- /dev/null
   +++ b/s/foobar
   @@ -0,0 +1,4 @@
   +woopwoop
   +
   +foo
   +bar
 
 check output when only diffing the subrepository
   $ hg diff -c . --subrepos s
   diff --git a/s/foobar b/s/foobar
   new file mode 100644
   index 0000000..8a5a5e2
   --- /dev/null
   +++ b/s/foobar
   @@ -0,0 +1,4 @@
   +woopwoop
   +
   +foo
   +bar
 
 check output when diffing something else
   $ hg diff -c . --subrepos .hgsubstate --nodates
   diff -r af6d2edbb0d3 -r 255ee8cf690e .hgsubstate
   --- a/.hgsubstate
   +++ b/.hgsubstate
   @@ -1,1 +1,1 @@
   -32a343883b74769118bb1d3b4b1fbf9156f4dddc s
   +fd4dbf828a5b2fcd36b2bcf21ea773820970d129 s
 
 add new changes, including whitespace
   $ cd s
   $ cat > foobar << EOF
   > woop    woop
   > 
   > foo
   > bar
   > EOF
   $ echo foo > barfoo
   $ git add barfoo
   $ cd ..
 
   $ hg diff --subrepos --ignore-all-space
   diff --git a/s/barfoo b/s/barfoo
   new file mode 100644
   index 0000000..257cc56
   --- /dev/null
   +++ b/s/barfoo
   @@ -0,0 +1* @@ (glob)
   +foo
   $ hg diff --subrepos s/foobar
   diff --git a/s/foobar b/s/foobar
   index 8a5a5e2..bd5812a 100644
   --- a/s/foobar
   +++ b/s/foobar
   @@ -1,4 +1,4 @@
   -woopwoop
   +woop    woop
    
    foo
    bar
 
 execute a diffstat
 the output contains a regex, because git 1.7.10 and 1.7.11
  change the amount of whitespace
   $ hg diff --subrepos --stat
   \s*barfoo \|\s+1 \+ (re)
   \s*foobar \|\s+2 \+- (re)
    2 files changed, 2 insertions\(\+\), 1 deletions?\(-\) (re)
 
 adding an include should ignore the other elements
   $ hg diff --subrepos -I s/foobar
   diff --git a/s/foobar b/s/foobar
   index 8a5a5e2..bd5812a 100644
   --- a/s/foobar
   +++ b/s/foobar
   @@ -1,4 +1,4 @@
   -woopwoop
   +woop    woop
    
    foo
    bar
 
 adding an exclude should ignore this element
   $ hg diff --subrepos -X s/foobar
   diff --git a/s/barfoo b/s/barfoo
   new file mode 100644
   index 0000000..257cc56
   --- /dev/null
   +++ b/s/barfoo
   @@ -0,0 +1* @@ (glob)
   +foo
 
 moving a file should show a removal and an add
   $ hg revert --all
   reverting subrepo ../gitroot
   $ cd s
   $ git mv foobar woop
   $ cd ..
   $ hg diff --subrepos
   diff --git a/s/foobar b/s/foobar
   deleted file mode 100644
   index 8a5a5e2..0000000
   --- a/s/foobar
   +++ /dev/null
   @@ -1,4 +0,0 @@
   -woopwoop
   -
   -foo
   -bar
   diff --git a/s/woop b/s/woop
   new file mode 100644
   index 0000000..8a5a5e2
   --- /dev/null
   +++ b/s/woop
   @@ -0,0 +1,4 @@
   +woopwoop
   +
   +foo
   +bar
   $ rm s/woop
 
 revert the subrepository
   $ hg revert --all
   reverting subrepo ../gitroot
 
   $ hg status --subrepos
   ? s/barfoo
   ? s/foobar.orig
 
   $ mv s/foobar.orig s/foobar
 
   $ hg revert --no-backup s
   reverting subrepo ../gitroot
 
   $ hg status --subrepos
   ? s/barfoo
 
 revert moves orig files to the right place
   $ echo 'bloop' > s/foobar
   $ hg revert --all --verbose --config 'ui.origbackuppath=.hg/origbackups'
   reverting subrepo ../gitroot
   creating directory: $TESTTMP/tc/.hg/origbackups/s
   saving current version of foobar as .hg/origbackups/s/foobar
   $ ls .hg/origbackups/s
   foobar
   $ rm -rf .hg/origbackups
 
 show file at specific revision
   $ cat > s/foobar << EOF
   > woop    woop
   > fooo bar
   > EOF
   $ hg commit --subrepos -m "updated foobar"
   committing subrepository s
   $ cat > s/foobar << EOF
   > current foobar
   > (should not be visible using hg cat)
   > EOF
 
   $ hg cat -r . s/foobar
   woop    woop
   fooo bar (no-eol)
   $ hg cat -r "parents(.)" s/foobar > catparents
 
   $ mkdir -p tmp/s
 
   $ hg cat -r "parents(.)" --output tmp/%% s/foobar
   $ diff tmp/% catparents
 
   $ hg cat -r "parents(.)" --output tmp/%s s/foobar
   $ diff tmp/foobar catparents
 
   $ hg cat -r "parents(.)" --output tmp/%d/otherfoobar s/foobar
   $ diff tmp/s/otherfoobar catparents
 
   $ hg cat -r "parents(.)" --output tmp/%p s/foobar
   $ diff tmp/s/foobar catparents
 
   $ hg cat -r "parents(.)" --output tmp/%H s/foobar
   $ diff tmp/255ee8cf690ec86e99b1e80147ea93ece117cd9d catparents
 
   $ hg cat -r "parents(.)" --output tmp/%R s/foobar
   $ diff tmp/10 catparents
 
   $ hg cat -r "parents(.)" --output tmp/%h s/foobar
   $ diff tmp/255ee8cf690e catparents
 
   $ rm tmp/10
   $ hg cat -r "parents(.)" --output tmp/%r s/foobar
   $ diff tmp/10 catparents
 
   $ mkdir tmp/tc
   $ hg cat -r "parents(.)" --output tmp/%b/foobar s/foobar
   $ diff tmp/tc/foobar catparents
 
 cleanup
   $ rm -r tmp
   $ rm catparents
 
 add git files, using either files or patterns
   $ echo "hsss! hsssssssh!" > s/snake.python
   $ echo "ccc" > s/c.c
   $ echo "cpp" > s/cpp.cpp
 
   $ hg add s/snake.python s/c.c s/cpp.cpp
   $ hg st --subrepos s
   M s/foobar
   A s/c.c
   A s/cpp.cpp
   A s/snake.python
   ? s/barfoo
   $ hg revert s
   reverting subrepo ../gitroot
 
   $ hg add --subrepos "glob:**.python"
   adding s/snake.python
   $ hg st --subrepos s
   A s/snake.python
   ? s/barfoo
   ? s/c.c
   ? s/cpp.cpp
   ? s/foobar.orig
   $ hg revert s
   reverting subrepo ../gitroot
 
   $ hg add --subrepos s
   adding s/barfoo
   adding s/c.c
   adding s/cpp.cpp
   adding s/foobar.orig
   adding s/snake.python
   $ hg st --subrepos s
   A s/barfoo
   A s/c.c
   A s/cpp.cpp
   A s/foobar.orig
   A s/snake.python
   $ hg revert s
   reverting subrepo ../gitroot
 make sure everything is reverted correctly
   $ hg st --subrepos s
   ? s/barfoo
   ? s/c.c
   ? s/cpp.cpp
   ? s/foobar.orig
   ? s/snake.python
 
   $ hg add --subrepos --exclude "path:s/c.c"
   adding s/barfoo
   adding s/cpp.cpp
   adding s/foobar.orig
   adding s/snake.python
   $ hg st --subrepos s
   A s/barfoo
   A s/cpp.cpp
   A s/foobar.orig
   A s/snake.python
   ? s/c.c
   $ hg revert --all -q
 
 .hgignore should not have influence in subrepos
   $ cat > .hgignore << EOF
   > syntax: glob
   > *.python
   > EOF
   $ hg add .hgignore
   $ hg add --subrepos "glob:**.python" s/barfoo
   adding s/snake.python
   $ hg st --subrepos s
   A s/barfoo
   A s/snake.python
   ? s/c.c
   ? s/cpp.cpp
   ? s/foobar.orig
   $ hg revert --all -q
 
 .gitignore should have influence,
 except for explicitly added files (no patterns)
   $ cat > s/.gitignore << EOF
   > *.python
   > EOF
   $ hg add s/.gitignore
   $ hg st --subrepos s
   A s/.gitignore
   ? s/barfoo
   ? s/c.c
   ? s/cpp.cpp
   ? s/foobar.orig
   $ hg st --subrepos s --all
   A s/.gitignore
   ? s/barfoo
   ? s/c.c
   ? s/cpp.cpp
   ? s/foobar.orig
   I s/snake.python
   C s/f
   C s/foobar
   C s/g
   $ hg add --subrepos "glob:**.python"
   $ hg st --subrepos s
   A s/.gitignore
   ? s/barfoo
   ? s/c.c
   ? s/cpp.cpp
   ? s/foobar.orig
   $ hg add --subrepos s/snake.python
   $ hg st --subrepos s
   A s/.gitignore
   A s/snake.python
   ? s/barfoo
   ? s/c.c
   ? s/cpp.cpp
   ? s/foobar.orig
 
 correctly do a dry run
   $ hg add --subrepos s --dry-run
   adding s/barfoo
   adding s/c.c
   adding s/cpp.cpp
   adding s/foobar.orig
   $ hg st --subrepos s
   A s/.gitignore
   A s/snake.python
   ? s/barfoo
   ? s/c.c
   ? s/cpp.cpp
   ? s/foobar.orig
 
 error given when adding an already tracked file
   $ hg add s/.gitignore
   s/.gitignore already tracked!
   [1]
   $ hg add s/g
   s/g already tracked!
   [1]
 
 removed files can be re-added
 removing files using 'rm' or 'git rm' has the same effect,
 since we ignore the staging area
   $ hg ci --subrepos -m 'snake'
   committing subrepository s
   $ cd s
   $ rm snake.python
 (remove leftover .hg so Mercurial doesn't look for a root here)
   $ rm -rf .hg
   $ hg status --subrepos --all .
   R snake.python
   ? barfoo
   ? c.c
   ? cpp.cpp
   ? foobar.orig
   C .gitignore
   C f
   C foobar
   C g
   $ git rm snake.python
   rm 'snake.python'
   $ hg status --subrepos --all .
   R snake.python
   ? barfoo
   ? c.c
   ? cpp.cpp
   ? foobar.orig
   C .gitignore
   C f
   C foobar
   C g
   $ touch snake.python
   $ cd ..
   $ hg add s/snake.python
   $ hg status -S
   M s/snake.python
   ? .hgignore
   ? s/barfoo
   ? s/c.c
   ? s/cpp.cpp
   ? s/foobar.orig
   $ hg revert --all -q
 
 make sure we show changed files, rather than changed subtrees
   $ mkdir s/foo
   $ touch s/foo/bwuh
   $ hg add s/foo/bwuh
   $ hg commit -S -m "add bwuh"
   committing subrepository s
   $ hg status -S --change .
   M .hgsubstate
   A s/foo/bwuh
   ? s/barfoo
   ? s/c.c
   ? s/cpp.cpp
   ? s/foobar.orig
   ? s/snake.python.orig
 
 #if git19
 
 test for Git CVE-2016-3068
   $ hg init malicious-subrepository
   $ cd malicious-subrepository
   $ echo "s = [git]ext::sh -c echo% pwned:% \$PWNED_MSG% >pwned.txt" > .hgsub
   $ git init s
   Initialized empty Git repository in $TESTTMP/tc/malicious-subrepository/s/.git/
   $ cd s
   $ git commit --allow-empty -m 'empty'
   [master (root-commit) 153f934] empty
   $ cd ..
   $ hg add .hgsub
   $ hg commit -m "add subrepo"
   $ cd ..
   $ rm -f pwned.txt
   $ unset GIT_ALLOW_PROTOCOL
   $ PWNED_MSG="your git is too old or mercurial has regressed" hg clone \
   > malicious-subrepository malicious-subrepository-protected
   Cloning into '$TESTTMP/tc/malicious-subrepository-protected/s'...
   fatal: transport 'ext' not allowed
   updating to branch default
   cloning subrepo s from ext::sh -c echo% pwned:% $PWNED_MSG% >pwned.txt
   abort: git clone error 128 in s (in subrepository "s")
   [255]
   $ f -Dq pwned.txt
   pwned.txt: file not found
 
 whitelisting of ext should be respected (that's the git submodule behaviour)
   $ rm -f pwned.txt
   $ env GIT_ALLOW_PROTOCOL=ext PWNED_MSG="you asked for it" hg clone \
   > malicious-subrepository malicious-subrepository-clone-allowed
   Cloning into '$TESTTMP/tc/malicious-subrepository-clone-allowed/s'...
   fatal: Could not read from remote repository.
   
   Please make sure you have the correct access rights
   and the repository exists.
   updating to branch default
   cloning subrepo s from ext::sh -c echo% pwned:% $PWNED_MSG% >pwned.txt
   abort: git clone error 128 in s (in subrepository "s")
   [255]
   $ f -Dq pwned.txt
   pwned: you asked for it
 
 #endif
 
 test for ssh exploit with git subrepos 2017-07-25
 
   $ hg init malicious-proxycommand
   $ cd malicious-proxycommand
   $ echo 's = [git]ssh://-oProxyCommand=rm${IFS}non-existent/path' > .hgsub
   $ git init s
   Initialized empty Git repository in $TESTTMP/tc/malicious-proxycommand/s/.git/
   $ cd s
   $ git commit --allow-empty -m 'empty'
   [master (root-commit) 153f934] empty
   $ cd ..
   $ hg add .hgsub
   $ hg ci -m 'add subrepo'
   $ cd ..
   $ hg clone malicious-proxycommand malicious-proxycommand-clone
   updating to branch default
   abort: potentially unsafe url: 'ssh://-oProxyCommand=rm${IFS}non-existent/path' (in subrepository "s")
   [255]
 
 also check that a percent encoded '-' (%2D) doesn't work
 
   $ cd malicious-proxycommand
   $ echo 's = [git]ssh://%2DoProxyCommand=rm${IFS}non-existent/path' > .hgsub
   $ hg ci -m 'change url to percent encoded'
   $ cd ..
   $ rm -r malicious-proxycommand-clone
   $ hg clone malicious-proxycommand malicious-proxycommand-clone
   updating to branch default
   abort: potentially unsafe url: 'ssh://-oProxyCommand=rm${IFS}non-existent/path' (in subrepository "s")
   [255]