largefiles: don't copy largefiles from working dir to the store while converting

Previously, if one or more largefiles for a repo being converted were not in the
usercache, the convert would abort with a reference to the largefile being
missing (as opposed to the previous patch, where the standin was referenced as
missing).  This is because commitctx() tries to copy all largefiles to the
local store, first from the user cache, and if the file isn't found there, from
the working directory.  No files will exist in the working directory during a
convert, however.  It is not sufficient to force the source repo to be local
before proceeding, because clone and pull do not download largefiles by default.

This is slightly less than ideal because while the conversion will now complete,
it won't be possible to update to revs with missing largefiles unless the user
intervenes manually, because there is no default path pointing back to the
source repo.  Ideally these files would be cached during the conversion.

This check could have been done in reposetup.commitctx() instead, but this
ensures the local store directory is created, which is necessary to enable the
standin matcher.

The rm -> 'rm -f' change in the test is to temporarily suppress an error
clearing the cache- as noted, the cache is is not repopulated during convert.
When that is fixed, this can be changed back and the verification errors will
disappear too.

hgext/largefiles/lfutil.py

@@ -1,469 +1,469
 # 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.
 
 '''largefiles utility code: must not import other modules in this package.'''
 
 import os
 import errno
 import platform
 import shutil
 import stat
 
 from mercurial import dirstate, httpconnection, match as match_, util, scmutil
 from mercurial.i18n import _
 
 shortname = '.hglf'
 longname = 'largefiles'
 
 
 # -- Portability wrappers ----------------------------------------------
 
 def dirstatewalk(dirstate, matcher, unknown=False, ignored=False):
     return dirstate.walk(matcher, [], unknown, ignored)
 
 def repoadd(repo, list):
     add = repo[None].add
     return add(list)
 
 def reporemove(repo, list, unlink=False):
     def remove(list, unlink):
         wlock = repo.wlock()
         try:
             if unlink:
                 for f in list:
                     try:
                         util.unlinkpath(repo.wjoin(f))
                     except OSError, inst:
                         if inst.errno != errno.ENOENT:
                             raise
             repo[None].forget(list)
         finally:
             wlock.release()
     return remove(list, unlink=unlink)
 
 def repoforget(repo, list):
     forget = repo[None].forget
     return forget(list)
 
 def findoutgoing(repo, remote, force):
     from mercurial import discovery
     outgoing = discovery.findcommonoutgoing(repo, remote.peer(), force=force)
     return outgoing.missing
 
 # -- Private worker functions ------------------------------------------
 
 def getminsize(ui, assumelfiles, opt, default=10):
     lfsize = opt
     if not lfsize and assumelfiles:
         lfsize = ui.config(longname, 'minsize', default=default)
     if lfsize:
         try:
             lfsize = float(lfsize)
         except ValueError:
             raise util.Abort(_('largefiles: size must be number (not %s)\n')
                              % lfsize)
     if lfsize is None:
         raise util.Abort(_('minimum size for largefiles must be specified'))
     return lfsize
 
 def link(src, dest):
     try:
         util.oslink(src, dest)
     except OSError:
         # if hardlinks fail, fallback on atomic copy
         dst = util.atomictempfile(dest)
         for chunk in util.filechunkiter(open(src, 'rb')):
             dst.write(chunk)
         dst.close()
         os.chmod(dest, os.stat(src).st_mode)
 
 def usercachepath(ui, hash):
     path = ui.configpath(longname, 'usercache', None)
     if path:
         path = os.path.join(path, hash)
     else:
         if os.name == 'nt':
             appdata = os.getenv('LOCALAPPDATA', os.getenv('APPDATA'))
             if appdata:
                 path = os.path.join(appdata, longname, hash)
         elif platform.system() == 'Darwin':
             home = os.getenv('HOME')
             if home:
                 path = os.path.join(home, 'Library', 'Caches',
                                     longname, hash)
         elif os.name == 'posix':
             path = os.getenv('XDG_CACHE_HOME')
             if path:
                 path = os.path.join(path, longname, hash)
             else:
                 home = os.getenv('HOME')
                 if home:
                     path = os.path.join(home, '.cache', longname, hash)
         else:
             raise util.Abort(_('unknown operating system: %s\n') % os.name)
     return path
 
 def inusercache(ui, hash):
     path = usercachepath(ui, hash)
     return path and os.path.exists(path)
 
 def findfile(repo, hash):
     if instore(repo, hash):
         repo.ui.note(_('found %s in store\n') % hash)
         return storepath(repo, hash)
     elif inusercache(repo.ui, hash):
         repo.ui.note(_('found %s in system cache\n') % hash)
         path = storepath(repo, hash)
         util.makedirs(os.path.dirname(path))
         link(usercachepath(repo.ui, hash), path)
         return path
     return None
 
 class largefilesdirstate(dirstate.dirstate):
     def __getitem__(self, key):
         return super(largefilesdirstate, self).__getitem__(unixpath(key))
     def normal(self, f):
         return super(largefilesdirstate, self).normal(unixpath(f))
     def remove(self, f):
         return super(largefilesdirstate, self).remove(unixpath(f))
     def add(self, f):
         return super(largefilesdirstate, self).add(unixpath(f))
     def drop(self, f):
         return super(largefilesdirstate, self).drop(unixpath(f))
     def forget(self, f):
         return super(largefilesdirstate, self).forget(unixpath(f))
     def normallookup(self, f):
         return super(largefilesdirstate, self).normallookup(unixpath(f))
 
 def openlfdirstate(ui, repo, create=True):
     '''
     Return a dirstate object that tracks largefiles: i.e. its root is
     the repo root, but it is saved in .hg/largefiles/dirstate.
     '''
     admin = repo.join(longname)
     opener = scmutil.opener(admin)
     lfdirstate = largefilesdirstate(opener, ui, repo.root,
                                      repo.dirstate._validate)
 
     # If the largefiles dirstate does not exist, populate and create
     # it. This ensures that we create it on the first meaningful
     # largefiles operation in a new clone.
     if create and not os.path.exists(os.path.join(admin, 'dirstate')):
         util.makedirs(admin)
         matcher = getstandinmatcher(repo)
         for standin in dirstatewalk(repo.dirstate, matcher):
             lfile = splitstandin(standin)
             hash = readstandin(repo, lfile)
             lfdirstate.normallookup(lfile)
             try:
                 if hash == hashfile(repo.wjoin(lfile)):
                     lfdirstate.normal(lfile)
             except OSError, err:
                 if err.errno != errno.ENOENT:
                     raise
     return lfdirstate
 
 def lfdirstatestatus(lfdirstate, repo, rev):
     match = match_.always(repo.root, repo.getcwd())
     s = lfdirstate.status(match, [], False, False, False)
     unsure, modified, added, removed, missing, unknown, ignored, clean = s
     for lfile in unsure:
         if repo[rev][standin(lfile)].data().strip() != \
                 hashfile(repo.wjoin(lfile)):
             modified.append(lfile)
         else:
             clean.append(lfile)
             lfdirstate.normal(lfile)
     return (modified, added, removed, missing, unknown, ignored, clean)
 
 def listlfiles(repo, rev=None, matcher=None):
     '''return a list of largefiles in the working copy or the
     specified changeset'''
 
     if matcher is None:
         matcher = getstandinmatcher(repo)
 
     # ignore unknown files in working directory
     return [splitstandin(f)
             for f in repo[rev].walk(matcher)
             if rev is not None or repo.dirstate[f] != '?']
 
 def instore(repo, hash):
     return os.path.exists(storepath(repo, hash))
 
 def storepath(repo, hash):
     return repo.join(os.path.join(longname, hash))
 
 def copyfromcache(repo, hash, filename):
     '''Copy the specified largefile from the repo or system cache to
     filename in the repository. Return true on success or false if the
     file was not found in either cache (which should not happened:
     this is meant to be called only after ensuring that the needed
     largefile exists in the cache).'''
     path = findfile(repo, hash)
     if path is None:
         return False
     util.makedirs(os.path.dirname(repo.wjoin(filename)))
     # The write may fail before the file is fully written, but we
     # don't use atomic writes in the working copy.
     shutil.copy(path, repo.wjoin(filename))
     return True
 
 def copytostore(repo, rev, file, uploaded=False):
     hash = readstandin(repo, file, rev)
     if instore(repo, hash):
         return
     copytostoreabsolute(repo, repo.wjoin(file), hash)
 
 def copyalltostore(repo, node):
     '''Copy all largefiles in a given revision to the store'''
 
     ctx = repo[node]
     for filename in ctx.files():
         if isstandin(filename) and filename in ctx.manifest():
             realfile = splitstandin(filename)
             copytostore(repo, ctx.node(), realfile)
 
 
 def copytostoreabsolute(repo, file, hash):
     util.makedirs(os.path.dirname(storepath(repo, hash)))
     if inusercache(repo.ui, hash):
         link(usercachepath(repo.ui, hash), storepath(repo, hash))
-    else:
+    elif not getattr(repo, "_isconverting", False):
         dst = util.atomictempfile(storepath(repo, hash),
                                   createmode=repo.store.createmode)
         for chunk in util.filechunkiter(open(file, 'rb')):
             dst.write(chunk)
         dst.close()
         linktousercache(repo, hash)
 
 def linktousercache(repo, hash):
     path = usercachepath(repo.ui, hash)
     if path:
         util.makedirs(os.path.dirname(path))
         link(storepath(repo, hash), path)
 
 def getstandinmatcher(repo, pats=[], opts={}):
     '''Return a match object that applies pats to the standin directory'''
     standindir = repo.pathto(shortname)
     if pats:
         # patterns supplied: search standin directory relative to current dir
         cwd = repo.getcwd()
         if os.path.isabs(cwd):
             # cwd is an absolute path for hg -R <reponame>
             # work relative to the repository root in this case
             cwd = ''
         pats = [os.path.join(standindir, cwd, pat) for pat in pats]
     elif os.path.isdir(standindir):
         # no patterns: relative to repo root
         pats = [standindir]
     else:
         # no patterns and no standin dir: return matcher that matches nothing
         match = match_.match(repo.root, None, [], exact=True)
         match.matchfn = lambda f: False
         return match
     return getmatcher(repo, pats, opts, showbad=False)
 
 def getmatcher(repo, pats=[], opts={}, showbad=True):
     '''Wrapper around scmutil.match() that adds showbad: if false,
     neuter the match object's bad() method so it does not print any
     warnings about missing files or directories.'''
     match = scmutil.match(repo[None], pats, opts)
 
     if not showbad:
         match.bad = lambda f, msg: None
     return match
 
 def composestandinmatcher(repo, rmatcher):
     '''Return a matcher that accepts standins corresponding to the
     files accepted by rmatcher. Pass the list of files in the matcher
     as the paths specified by the user.'''
     smatcher = getstandinmatcher(repo, rmatcher.files())
     isstandin = smatcher.matchfn
     def composedmatchfn(f):
         return isstandin(f) and rmatcher.matchfn(splitstandin(f))
     smatcher.matchfn = composedmatchfn
 
     return smatcher
 
 def standin(filename):
     '''Return the repo-relative path to the standin for the specified big
     file.'''
     # Notes:
     # 1) Some callers want an absolute path, but for instance addlargefiles
     #    needs it repo-relative so it can be passed to repoadd().  So leave
     #    it up to the caller to use repo.wjoin() to get an absolute path.
     # 2) Join with '/' because that's what dirstate always uses, even on
     #    Windows. Change existing separator to '/' first in case we are
     #    passed filenames from an external source (like the command line).
     return shortname + '/' + util.pconvert(filename)
 
 def isstandin(filename):
     '''Return true if filename is a big file standin. filename must be
     in Mercurial's internal form (slash-separated).'''
     return filename.startswith(shortname + '/')
 
 def splitstandin(filename):
     # Split on / because that's what dirstate always uses, even on Windows.
     # Change local separator to / first just in case we are passed filenames
     # from an external source (like the command line).
     bits = util.pconvert(filename).split('/', 1)
     if len(bits) == 2 and bits[0] == shortname:
         return bits[1]
     else:
         return None
 
 def updatestandin(repo, standin):
     file = repo.wjoin(splitstandin(standin))
     if os.path.exists(file):
         hash = hashfile(file)
         executable = getexecutable(file)
         writestandin(repo, standin, hash, executable)
 
 def readstandin(repo, filename, node=None):
     '''read hex hash from standin for filename at given node, or working
     directory if no node is given'''
     return repo[node][standin(filename)].data().strip()
 
 def writestandin(repo, standin, hash, executable):
     '''write hash to <repo.root>/<standin>'''
     writehash(hash, repo.wjoin(standin), executable)
 
 def copyandhash(instream, outfile):
     '''Read bytes from instream (iterable) and write them to outfile,
     computing the SHA-1 hash of the data along the way.  Close outfile
     when done and return the binary hash.'''
     hasher = util.sha1('')
     for data in instream:
         hasher.update(data)
         outfile.write(data)
 
     # Blecch: closing a file that somebody else opened is rude and
     # wrong. But it's so darn convenient and practical! After all,
     # outfile was opened just to copy and hash.
     outfile.close()
 
     return hasher.digest()
 
 def hashrepofile(repo, file):
     return hashfile(repo.wjoin(file))
 
 def hashfile(file):
     if not os.path.exists(file):
         return ''
     hasher = util.sha1('')
     fd = open(file, 'rb')
     for data in blockstream(fd):
         hasher.update(data)
     fd.close()
     return hasher.hexdigest()
 
 class limitreader(object):
     def __init__(self, f, limit):
         self.f = f
         self.limit = limit
 
     def read(self, length):
         if self.limit == 0:
             return ''
         length = length > self.limit and self.limit or length
         self.limit -= length
         return self.f.read(length)
 
     def close(self):
         pass
 
 def blockstream(infile, blocksize=128 * 1024):
     """Generator that yields blocks of data from infile and closes infile."""
     while True:
         data = infile.read(blocksize)
         if not data:
             break
         yield data
     # same blecch as copyandhash() above
     infile.close()
 
 def writehash(hash, filename, executable):
     util.makedirs(os.path.dirname(filename))
     util.writefile(filename, hash + '\n')
     os.chmod(filename, getmode(executable))
 
 def getexecutable(filename):
     mode = os.stat(filename).st_mode
     return ((mode & stat.S_IXUSR) and
             (mode & stat.S_IXGRP) and
             (mode & stat.S_IXOTH))
 
 def getmode(executable):
     if executable:
         return 0755
     else:
         return 0644
 
 def urljoin(first, second, *arg):
     def join(left, right):
         if not left.endswith('/'):
             left += '/'
         if right.startswith('/'):
             right = right[1:]
         return left + right
 
     url = join(first, second)
     for a in arg:
         url = join(url, a)
     return url
 
 def hexsha1(data):
     """hexsha1 returns the hex-encoded sha1 sum of the data in the file-like
     object data"""
     h = util.sha1()
     for chunk in util.filechunkiter(data):
         h.update(chunk)
     return h.hexdigest()
 
 def httpsendfile(ui, filename):
     return httpconnection.httpsendfile(ui, filename, 'rb')
 
 def unixpath(path):
     '''Return a version of path normalized for use with the lfdirstate.'''
     return util.pconvert(os.path.normpath(path))
 
 def islfilesrepo(repo):
     if ('largefiles' in repo.requirements and
             util.any(shortname + '/' in f[0] for f in repo.store.datafiles())):
         return True
 
     return util.any(openlfdirstate(repo.ui, repo, False))
 
 class storeprotonotcapable(Exception):
     def __init__(self, storetypes):
         self.storetypes = storetypes
 
 def getcurrentheads(repo):
     branches = repo.branchmap()
     heads = []
     for branch in branches:
         newheads = repo.branchheads(branch)
         heads = heads + newheads
     return heads
 
 def getstandinsstate(repo):
     standins = []
     matcher = getstandinmatcher(repo)
     for standin in dirstatewalk(repo.dirstate, matcher):
         lfile = splitstandin(standin)
         standins.append((lfile, readstandin(repo, lfile)))
     return standins
 
 def getlfilestoupdate(oldstandins, newstandins):
     changedstandins = set(oldstandins).symmetric_difference(set(newstandins))
     filelist = []
     for f in changedstandins:
         if f[0] not in filelist:
             filelist.append(f[0])
     return filelist

hgext/largefiles/overrides.py

@@ -1,1115 +1,1123
 # Copyright 2009-2010 Gregory P. Ward
 # Copyright 2009-2010 Intelerad Medical Systems Incorporated
 # Copyright 2010-2011 Fog Creek Software
 # Copyright 2010-2011 Unity Technologies
 #
 # This software may be used and distributed according to the terms of the
 # GNU General Public License version 2 or any later version.
 
 '''Overridden Mercurial commands and functions for the largefiles extension'''
 
 import os
 import copy
 
 from mercurial import hg, commands, util, cmdutil, scmutil, match as match_, \
     node, archival, error, merge
 from mercurial.i18n import _
 from mercurial.node import hex
 from hgext import rebase
 
 import lfutil
 import lfcommands
 
 # -- Utility functions: commonly/repeatedly needed functionality ---------------
 
 def installnormalfilesmatchfn(manifest):
     '''overrides scmutil.match so that the matcher it returns will ignore all
     largefiles'''
     oldmatch = None # for the closure
     def overridematch(ctx, pats=[], opts={}, globbed=False,
             default='relpath'):
         match = oldmatch(ctx, pats, opts, globbed, default)
         m = copy.copy(match)
         notlfile = lambda f: not (lfutil.isstandin(f) or lfutil.standin(f) in
                 manifest)
         m._files = filter(notlfile, m._files)
         m._fmap = set(m._files)
         origmatchfn = m.matchfn
         m.matchfn = lambda f: notlfile(f) and origmatchfn(f) or None
         return m
     oldmatch = installmatchfn(overridematch)
 
 def installmatchfn(f):
     oldmatch = scmutil.match
     setattr(f, 'oldmatch', oldmatch)
     scmutil.match = f
     return oldmatch
 
 def restorematchfn():
     '''restores scmutil.match to what it was before installnormalfilesmatchfn
     was called.  no-op if scmutil.match is its original function.
 
     Note that n calls to installnormalfilesmatchfn will require n calls to
     restore matchfn to reverse'''
     scmutil.match = getattr(scmutil.match, 'oldmatch', scmutil.match)
 
 def addlargefiles(ui, repo, *pats, **opts):
     large = opts.pop('large', None)
     lfsize = lfutil.getminsize(
         ui, lfutil.islfilesrepo(repo), opts.pop('lfsize', None))
 
     lfmatcher = None
     if lfutil.islfilesrepo(repo):
         lfpats = ui.configlist(lfutil.longname, 'patterns', default=[])
         if lfpats:
             lfmatcher = match_.match(repo.root, '', list(lfpats))
 
     lfnames = []
     m = scmutil.match(repo[None], pats, opts)
     m.bad = lambda x, y: None
     wctx = repo[None]
     for f in repo.walk(m):
         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(_('%s already a largefile\n') % f)
             continue
 
         if (exact or not exists) and not lfutil.isstandin(f):
             wfile = repo.wjoin(f)
 
             # In case the file was removed previously, but not committed
             # (issue3507)
             if not os.path.exists(wfile):
                 continue
 
             abovemin = (lfsize and
                         os.lstat(wfile).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(_('adding %s as a largefile\n') % m.rel(f))
 
     bad = []
     standins = []
 
     # Need to lock, otherwise there could be a race condition between
     # when standins are created and added to the repo.
     wlock = repo.wlock()
     try:
         if not opts.get('dry_run'):
             lfdirstate = lfutil.openlfdirstate(ui, repo)
             for f in lfnames:
                 standinname = lfutil.standin(f)
                 lfutil.writestandin(repo, standinname, hash='',
                     executable=lfutil.getexecutable(repo.wjoin(f)))
                 standins.append(standinname)
                 if lfdirstate[f] == 'r':
                     lfdirstate.normallookup(f)
                 else:
                     lfdirstate.add(f)
             lfdirstate.write()
             bad += [lfutil.splitstandin(f)
                     for f in lfutil.repoadd(repo, standins)
                     if f in m.files()]
     finally:
         wlock.release()
     return bad
 
 def removelargefiles(ui, repo, *pats, **opts):
     after = opts.get('after')
     if not pats and not after:
         raise util.Abort(_('no files specified'))
     m = scmutil.match(repo[None], pats, opts)
     try:
         repo.lfstatus = True
         s = repo.status(match=m, clean=True)
     finally:
         repo.lfstatus = False
     manifest = repo[None].manifest()
     modified, added, deleted, clean = [[f for f in list
                                         if lfutil.standin(f) in manifest]
                                        for list in [s[0], s[1], s[3], s[6]]]
 
     def warn(files, reason):
         for f in files:
             ui.warn(_('not removing %s: %s (use forget to undo)\n')
                     % (m.rel(f), reason))
         return int(len(files) > 0)
 
     result = 0
 
     if after:
         remove, forget = deleted, []
         result = warn(modified + added + clean, _('file still exists'))
     else:
         remove, forget = deleted + clean, []
         result = warn(modified, _('file is modified'))
         result = warn(added, _('file has been marked for add')) or result
 
     for f in sorted(remove + forget):
         if ui.verbose or not m.exact(f):
             ui.status(_('removing %s\n') % m.rel(f))
 
     # Need to lock because standin files are deleted then removed from the
     # repository and we could race in-between.
     wlock = repo.wlock()
     try:
         lfdirstate = lfutil.openlfdirstate(ui, repo)
         for f in remove:
             if not after:
                 # If this is being called by addremove, notify the user that we
                 # are removing the file.
                 if getattr(repo, "_isaddremove", False):
                     ui.status(_('removing %s\n') % f)
                 if os.path.exists(repo.wjoin(f)):
                     util.unlinkpath(repo.wjoin(f))
             lfdirstate.remove(f)
         lfdirstate.write()
         forget = [lfutil.standin(f) for f in forget]
         remove = [lfutil.standin(f) for f in remove]
         lfutil.repoforget(repo, forget)
         # If this is being called by addremove, let the original addremove
         # function handle this.
         if not getattr(repo, "_isaddremove", False):
             lfutil.reporemove(repo, remove, unlink=True)
         else:
             lfutil.reporemove(repo, remove, unlink=False)
     finally:
         wlock.release()
 
     return result
 
 # For overriding mercurial.hgweb.webcommands so that largefiles will
 # appear at their right place in the manifests.
 def decodepath(orig, path):
     return lfutil.splitstandin(path) or path
 
 # -- Wrappers: modify existing commands --------------------------------
 
 # Add works by going through the files that the user wanted to add and
 # checking if they should be added as largefiles. Then it makes a new
 # matcher which matches only the normal files and runs the original
 # version of add.
 def overrideadd(orig, ui, repo, *pats, **opts):
     normal = opts.pop('normal')
     if normal:
         if opts.get('large'):
             raise util.Abort(_('--normal cannot be used with --large'))
         return orig(ui, repo, *pats, **opts)
     bad = addlargefiles(ui, repo, *pats, **opts)
     installnormalfilesmatchfn(repo[None].manifest())
     result = orig(ui, repo, *pats, **opts)
     restorematchfn()
 
     return (result == 1 or bad) and 1 or 0
 
 def overrideremove(orig, ui, repo, *pats, **opts):
     installnormalfilesmatchfn(repo[None].manifest())
     result = orig(ui, repo, *pats, **opts)
     restorematchfn()
     return removelargefiles(ui, repo, *pats, **opts) or result
 
 def overridestatusfn(orig, repo, rev2, **opts):
     try:
         repo._repo.lfstatus = True
         return orig(repo, rev2, **opts)
     finally:
         repo._repo.lfstatus = False
 
 def overridestatus(orig, ui, repo, *pats, **opts):
     try:
         repo.lfstatus = True
         return orig(ui, repo, *pats, **opts)
     finally:
         repo.lfstatus = False
 
 def overridedirty(orig, repo, ignoreupdate=False):
     try:
         repo._repo.lfstatus = True
         return orig(repo, ignoreupdate)
     finally:
         repo._repo.lfstatus = False
 
 def overridelog(orig, ui, repo, *pats, **opts):
     try:
         repo.lfstatus = True
         return orig(ui, repo, *pats, **opts)
     finally:
         repo.lfstatus = False
 
 def overrideverify(orig, ui, repo, *pats, **opts):
     large = opts.pop('large', False)
     all = opts.pop('lfa', False)
     contents = opts.pop('lfc', False)
 
     result = orig(ui, repo, *pats, **opts)
     if large:
         result = result or lfcommands.verifylfiles(ui, repo, all, contents)
     return result
 
 # Override needs to refresh standins so that update's normal merge
 # will go through properly. Then the other update hook (overriding repo.update)
 # will get the new files. Filemerge is also overridden so that the merge
 # will merge standins correctly.
 def overrideupdate(orig, ui, repo, *pats, **opts):
     lfdirstate = lfutil.openlfdirstate(ui, repo)
     s = lfdirstate.status(match_.always(repo.root, repo.getcwd()), [], False,
         False, False)
     (unsure, modified, added, removed, missing, unknown, ignored, clean) = s
 
     # Need to lock between the standins getting updated and their
     # largefiles getting updated
     wlock = repo.wlock()
     try:
         if opts['check']:
             mod = len(modified) > 0
             for lfile in unsure:
                 standin = lfutil.standin(lfile)
                 if repo['.'][standin].data().strip() != \
                         lfutil.hashfile(repo.wjoin(lfile)):
                     mod = True
                 else:
                     lfdirstate.normal(lfile)
             lfdirstate.write()
             if mod:
                 raise util.Abort(_('uncommitted local changes'))
         # XXX handle removed differently
         if not opts['clean']:
             for lfile in unsure + modified + added:
                 lfutil.updatestandin(repo, lfutil.standin(lfile))
     finally:
         wlock.release()
     return orig(ui, repo, *pats, **opts)
 
 # Before starting the manifest merge, merge.updates will call
 # _checkunknown 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 manifestmerge function below.
 def overridecheckunknownfile(origfn, repo, wctx, mctx, f):
     if lfutil.standin(f) in wctx:
         return False
     return origfn(repo, wctx, mctx, f)
 
 # 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 manifestmerge 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.
 def overridemanifestmerge(origfn, repo, p1, p2, pa, overwrite, partial):
     actions = origfn(repo, p1, p2, pa, overwrite, partial)
     processed = []
 
     for action in actions:
         if overwrite:
             processed.append(action)
             continue
         f, m = action[:2]
 
         choices = (_('&Largefile'), _('&Normal file'))
         if m == "g" and lfutil.splitstandin(f) in p1 and f in p2:
             # Case 1: normal file in the working copy, largefile in
             # the second parent
             lfile = lfutil.splitstandin(f)
             standin = f
             msg = _('%s has been turned into a largefile\n'
                     'use (l)argefile or keep as (n)ormal file?') % lfile
             if repo.ui.promptchoice(msg, choices, 0) == 0:
                 processed.append((lfile, "r"))
                 processed.append((standin, "g", p2.flags(standin)))
             else:
                 processed.append((standin, "r"))
         elif m == "g" and lfutil.standin(f) in p1 and f in p2:
             # Case 2: largefile in the working copy, normal file in
             # the second parent
             standin = lfutil.standin(f)
             lfile = f
             msg = _('%s has been turned into a normal file\n'
                     'keep as (l)argefile or use (n)ormal file?') % lfile
             if repo.ui.promptchoice(msg, choices, 0) == 0:
                 processed.append((lfile, "r"))
             else:
                 processed.append((standin, "r"))
                 processed.append((lfile, "g", p2.flags(lfile)))
         else:
             processed.append(action)
 
     return processed
 
 # Override filemerge to prompt the user about how they wish to merge
 # largefiles. This will handle identical edits, and copy/rename +
 # edit without prompting the user.
 def overridefilemerge(origfn, repo, mynode, orig, fcd, fco, fca):
     # Use better variable names here. Because this is a wrapper we cannot
     # change the variable names in the function declaration.
     fcdest, fcother, fcancestor = fcd, fco, fca
     if not lfutil.isstandin(orig):
         return origfn(repo, mynode, orig, fcdest, fcother, fcancestor)
     else:
         if not fcother.cmp(fcdest): # files identical?
             return None
 
         # backwards, use working dir parent as ancestor
         if fcancestor == fcother:
             fcancestor = fcdest.parents()[0]
 
         if orig != fcother.path():
             repo.ui.status(_('merging %s and %s to %s\n')
                            % (lfutil.splitstandin(orig),
                               lfutil.splitstandin(fcother.path()),
                               lfutil.splitstandin(fcdest.path())))
         else:
             repo.ui.status(_('merging %s\n')
                            % lfutil.splitstandin(fcdest.path()))
 
         if fcancestor.path() != fcother.path() and fcother.data() == \
                 fcancestor.data():
             return 0
         if fcancestor.path() != fcdest.path() and fcdest.data() == \
                 fcancestor.data():
             repo.wwrite(fcdest.path(), fcother.data(), fcother.flags())
             return 0
 
         if repo.ui.promptchoice(_('largefile %s has a merge conflict\n'
                              'keep (l)ocal or take (o)ther?') %
                              lfutil.splitstandin(orig),
                              (_('&Local'), _('&Other')), 0) == 0:
             return 0
         else:
             repo.wwrite(fcdest.path(), fcother.data(), fcother.flags())
             return 0
 
 # 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.
 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)
 
     def makestandin(relpath):
         path = scmutil.canonpath(repo.root, repo.getcwd(), relpath)
         return os.path.join(repo.wjoin(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))
     # 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
     try:
         try:
             installnormalfilesmatchfn(repo[None].manifest())
             result = orig(ui, repo, pats, opts, rename)
         except util.Abort, e:
             if str(e) != _('no files to copy'):
                 raise e
             else:
                 nonormalfiles = True
             result = 0
     finally:
         restorematchfn()
 
     # 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
 
     try:
         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()
             oldmatch = None # for the closure
             def overridematch(ctx, pats=[], opts={}, globbed=False,
                     default='relpath'):
                 newpats = []
                 # The patterns were previously mangled to add the standin
                 # directory; we need to remove that now
                 for pat in pats:
                     if match_.patkind(pat) is None and lfutil.shortname in pat:
                         newpats.append(pat.replace(lfutil.shortname, ''))
                     else:
                         newpats.append(pat)
                 match = oldmatch(ctx, newpats, opts, globbed, default)
                 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._fmap = set(m._files)
                 origmatchfn = m.matchfn
                 m.matchfn = lambda f: (lfutil.isstandin(f) and
                                     (f in manifest) and
                                     origmatchfn(lfutil.splitstandin(f)) or
                                     None)
                 return m
             oldmatch = installmatchfn(overridematch)
             listpats = []
             for pat in pats:
                 if match_.patkind(pat) is not None:
                     listpats.append(pat)
                 else:
                     listpats.append(makestandin(pat))
 
             try:
                 origcopyfile = util.copyfile
                 copiedfiles = []
                 def overridecopyfile(src, dest):
                     if (lfutil.shortname in src and
                         dest.startswith(repo.wjoin(lfutil.shortname))):
                         destlfile = dest.replace(lfutil.shortname, '')
                         if not opts['force'] and os.path.exists(destlfile):
                             raise IOError('',
                                 _('destination largefile already exists'))
                     copiedfiles.append((src, dest))
                     origcopyfile(src, dest)
 
                 util.copyfile = overridecopyfile
                 result += orig(ui, repo, listpats, opts, rename)
             finally:
                 util.copyfile = origcopyfile
 
             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('')), '')
                     destlfile = dest.replace(repo.wjoin(lfutil.standin('')), '')
                     destlfiledir = os.path.dirname(repo.wjoin(destlfile)) or '.'
                     if not os.path.isdir(destlfiledir):
                         os.makedirs(destlfiledir)
                     if rename:
                         os.rename(repo.wjoin(srclfile), repo.wjoin(destlfile))
                         lfdirstate.remove(srclfile)
                     else:
                         util.copyfile(repo.wjoin(srclfile),
                                       repo.wjoin(destlfile))
 
                     lfdirstate.add(destlfile)
             lfdirstate.write()
         except util.Abort, e:
             if str(e) != _('no files to copy'):
                 raise e
             else:
                 nolfiles = True
     finally:
         restorematchfn()
         wlock.release()
 
     if nolfiles and nonormalfiles:
         raise util.Abort(_('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. Then return the standins
 # to their proper state
 def overriderevert(orig, ui, repo, *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.
     wlock = repo.wlock()
     try:
         lfdirstate = lfutil.openlfdirstate(ui, repo)
         (modified, added, removed, missing, unknown, ignored, clean) = \
             lfutil.lfdirstatestatus(lfdirstate, repo, repo['.'].rev())
         for lfile in modified:
             lfutil.updatestandin(repo, lfutil.standin(lfile))
         for lfile in missing:
             if (os.path.exists(repo.wjoin(lfutil.standin(lfile)))):
                 os.unlink(repo.wjoin(lfutil.standin(lfile)))
 
         try:
             ctx = scmutil.revsingle(repo, opts.get('rev'))
             oldmatch = None # for the closure
             def overridematch(ctx, pats=[], opts={}, globbed=False,
                     default='relpath'):
                 match = oldmatch(ctx, pats, opts, globbed, default)
                 m = copy.copy(match)
                 def tostandin(f):
                     if lfutil.standin(f) in ctx:
                         return lfutil.standin(f)
                     elif lfutil.standin(f) in repo[None]:
                         return None
                     return f
                 m._files = [tostandin(f) for f in m._files]
                 m._files = [f for f in m._files if f is not None]
                 m._fmap = set(m._files)
                 origmatchfn = m.matchfn
                 def matchfn(f):
                     if lfutil.isstandin(f):
                         # We need to keep track of what largefiles are being
                         # matched so we know which ones to update later --
                         # otherwise we accidentally revert changes to other
                         # largefiles. This is repo-specific, so duckpunch the
                         # repo object to keep the list of largefiles for us
                         # later.
                         if origmatchfn(lfutil.splitstandin(f)) and \
                                 (f in repo[None] or f in ctx):
                             lfileslist = getattr(repo, '_lfilestoupdate', [])
                             lfileslist.append(lfutil.splitstandin(f))
                             repo._lfilestoupdate = lfileslist
                             return True
                         else:
                             return False
                     return origmatchfn(f)
                 m.matchfn = matchfn
                 return m
             oldmatch = installmatchfn(overridematch)
             scmutil.match
             matches = overridematch(repo[None], pats, opts)
             orig(ui, repo, *pats, **opts)
         finally:
             restorematchfn()
         lfileslist = getattr(repo, '_lfilestoupdate', [])
         lfcommands.updatelfiles(ui, repo, filelist=lfileslist,
                                 printmessage=False)
 
         # empty out the largefiles list so we start fresh next time
         repo._lfilestoupdate = []
         for lfile in modified:
             if lfile in lfileslist:
                 if os.path.exists(repo.wjoin(lfutil.standin(lfile))) and lfile\
                         in repo['.']:
                     lfutil.writestandin(repo, lfutil.standin(lfile),
                         repo['.'][lfile].data().strip(),
                         'x' in repo['.'][lfile].flags())
         lfdirstate = lfutil.openlfdirstate(ui, repo)
         for lfile in added:
             standin = lfutil.standin(lfile)
             if standin not in ctx and (standin in matches or opts.get('all')):
                 if lfile in lfdirstate:
                     lfdirstate.drop(lfile)
                 util.unlinkpath(repo.wjoin(standin))
         lfdirstate.write()
     finally:
         wlock.release()
 
 def hgupdate(orig, repo, node):
     # Only call updatelfiles the standins that have changed to save time
     oldstandins = lfutil.getstandinsstate(repo)
     result = orig(repo, node)
     newstandins = lfutil.getstandinsstate(repo)
     filelist = lfutil.getlfilestoupdate(oldstandins, newstandins)
     lfcommands.updatelfiles(repo.ui, repo, filelist=filelist, printmessage=True)
     return result
 
 def hgclean(orig, repo, node, show_stats=True):
     result = orig(repo, node, show_stats)
     lfcommands.updatelfiles(repo.ui, repo)
     return result
 
 def hgmerge(orig, repo, node, force=None, remind=True):
     # Mark the repo as being in the middle of a merge, so that
     # updatelfiles() will know that it needs to trust the standins in
     # the working copy, not in the standins in the current node
     repo._ismerging = True
     try:
         result = orig(repo, node, force, remind)
         lfcommands.updatelfiles(repo.ui, repo)
     finally:
         repo._ismerging = False
     return result
 
 # When we rebase a repository with remotely changed largefiles, we need to
 # take some extra care so that the largefiles are correctly updated in the
 # working copy
 def overridepull(orig, ui, repo, source=None, **opts):
     revsprepull = len(repo)
     if opts.get('rebase', False):
         repo._isrebasing = True
         try:
             if opts.get('update'):
                 del opts['update']
                 ui.debug('--update and --rebase are not compatible, ignoring '
                          'the update flag\n')
             del opts['rebase']
             cmdutil.bailifchanged(repo)
             origpostincoming = commands.postincoming
             def _dummy(*args, **kwargs):
                 pass
             commands.postincoming = _dummy
             if not source:
                 source = 'default'
             repo.lfpullsource = source
             try:
                 result = commands.pull(ui, repo, source, **opts)
             finally:
                 commands.postincoming = origpostincoming
             revspostpull = len(repo)
             if revspostpull > revsprepull:
                 result = result or rebase.rebase(ui, repo)
         finally:
             repo._isrebasing = False
     else:
         if not source:
             source = 'default'
         repo.lfpullsource = source
         oldheads = lfutil.getcurrentheads(repo)
         result = orig(ui, repo, source, **opts)
         # If we do not have the new largefiles for any new heads we pulled, we
         # will run into a problem later if we try to merge or rebase with one of
         # these heads, so cache the largefiles now directly into the system
         # cache.
         ui.status(_("caching new largefiles\n"))
         numcached = 0
         heads = lfutil.getcurrentheads(repo)
         newheads = set(heads).difference(set(oldheads))
         for head in newheads:
             (cached, missing) = lfcommands.cachelfiles(ui, repo, head)
             numcached += len(cached)
         ui.status(_("%d largefiles cached\n") % numcached)
     if opts.get('all_largefiles'):
         revspostpull = len(repo)
         revs = []
         for rev in xrange(revsprepull + 1, revspostpull):
             revs.append(repo[rev].rev())
         lfcommands.downloadlfiles(ui, repo, revs)
     return result
 
 def overrideclone(orig, ui, source, dest=None, **opts):
     d = dest
     if d is None:
         d = hg.defaultdest(source)
     if opts.get('all_largefiles') and not hg.islocal(d):
             raise util.Abort(_(
             '--all-largefiles is incompatible with non-local destination %s' %
             d))
 
     return orig(ui, source, dest, **opts)
 
 def hgclone(orig, ui, opts, *args, **kwargs):
     result = orig(ui, opts, *args, **kwargs)
 
     if result is not None:
         sourcerepo, destrepo = result
         repo = destrepo.local()
 
         # The .hglf directory must exist for the standin matcher to match
         # anything (which listlfiles uses for each rev), and .hg/largefiles is
         # assumed to exist by the code that caches the downloaded file.  These
         # directories exist if clone updated to any rev.  (If the repo does not
         # have largefiles, download never gets to the point of needing
         # .hg/largefiles, and the standin matcher won't match anything anyway.)
         if 'largefiles' in repo.requirements:
             if opts.get('noupdate'):
                 util.makedirs(repo.pathto(lfutil.shortname))
                 util.makedirs(repo.join(lfutil.longname))
 
         # 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('all_largefiles'):
             success, missing = lfcommands.downloadlfiles(ui, repo, None)
 
             if missing != 0:
                 return None
 
     return result
 
 def overriderebase(orig, ui, repo, **opts):
     repo._isrebasing = True
     try:
         return orig(ui, repo, **opts)
     finally:
         repo._isrebasing = False
 
 def overridearchive(orig, repo, dest, node, kind, decode=True, matchfn=None,
             prefix=None, mtime=None, subrepos=None):
     # No need to lock because we are only reading history and
     # largefile caches, neither of which are modified.
     lfcommands.cachelfiles(repo.ui, repo, node)
 
     if kind not in archival.archivers:
         raise util.Abort(_("unknown archive type '%s'") % kind)
 
     ctx = repo[node]
 
     if kind == 'files':
         if prefix:
             raise util.Abort(
                 _('cannot give prefix when archiving to files'))
     else:
         prefix = archival.tidyprefix(dest, kind, prefix)
 
     def write(name, mode, islink, getdata):
         if matchfn and not matchfn(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("ui", "archivemeta", True):
         def metadata():
             base = 'repo: %s\nnode: %s\nbranch: %s\n' % (
                 hex(repo.changelog.node(0)), hex(node), ctx.branch())
 
             tags = ''.join('tag: %s\n' % t for t in ctx.tags()
                            if repo.tagtype(t) == 'global')
             if not tags:
                 repo.ui.pushbuffer()
                 opts = {'template': '{latesttag}\n{latesttagdistance}',
                         'style': '', 'patch': None, 'git': None}
                 cmdutil.show_changeset(repo.ui, repo, opts).show(ctx)
                 ltags, dist = repo.ui.popbuffer().split('\n')
                 tags = ''.join('latesttag: %s\n' % t for t in ltags.split(':'))
                 tags += 'latesttagdistance: %s\n' % dist
 
             return base + tags
 
         write('.hg_archival.txt', 0644, False, metadata)
 
     for f in ctx:
         ff = ctx.flags(f)
         getdata = ctx[f].data
         if lfutil.isstandin(f):
             path = lfutil.findfile(repo, getdata().strip())
             if path is None:
                 raise util.Abort(
                     _('largefile %s not found in repo store or system cache')
                     % lfutil.splitstandin(f))
             f = lfutil.splitstandin(f)
 
             def getdatafn():
                 fd = None
                 try:
                     fd = open(path, 'rb')
                     return fd.read()
                 finally:
                     if fd:
                         fd.close()
 
             getdata = getdatafn
         write(f, 'x' in ff and 0755 or 0644, 'l' in ff, getdata)
 
     if subrepos:
         for subpath in ctx.substate:
             sub = ctx.sub(subpath)
             submatch = match_.narrowmatcher(subpath, matchfn)
             sub.archive(repo.ui, archiver, prefix, submatch)
 
     archiver.done()
 
 def hgsubrepoarchive(orig, repo, ui, archiver, prefix, match=None):
     repo._get(repo._state + ('hg',))
     rev = repo._state[1]
     ctx = repo._repo[rev]
 
     lfcommands.cachelfiles(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()
 
         archiver.addfile(prefix + repo._path + '/' + name, mode, islink, data)
 
     for f in ctx:
         ff = ctx.flags(f)
         getdata = ctx[f].data
         if lfutil.isstandin(f):
             path = lfutil.findfile(repo._repo, getdata().strip())
             if path is None:
                 raise util.Abort(
                     _('largefile %s not found in repo store or system cache')
                     % lfutil.splitstandin(f))
             f = lfutil.splitstandin(f)
 
             def getdatafn():
                 fd = None
                 try:
                     fd = open(os.path.join(prefix, path), 'rb')
                     return fd.read()
                 finally:
                     if fd:
                         fd.close()
 
             getdata = getdatafn
 
         write(f, 'x' in ff and 0755 or 0644, 'l' in ff, getdata)
 
     for subpath in ctx.substate:
         sub = ctx.sub(subpath)
         submatch = match_.narrowmatcher(subpath, match)
         sub.archive(ui, archiver, os.path.join(prefix, repo._path) + '/',
                     submatch)
 
 # 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 and backout.
 def overridebailifchanged(orig, repo):
     orig(repo)
     repo.lfstatus = True
     modified, added, removed, deleted = repo.status()[:4]
     repo.lfstatus = False
     if modified or added or removed or deleted:
         raise util.Abort(_('outstanding uncommitted changes'))
 
 # Fetch doesn't use cmdutil.bailifchanged so override it to add the check
 def overridefetch(orig, ui, repo, *pats, **opts):
     repo.lfstatus = True
     modified, added, removed, deleted = repo.status()[:4]
     repo.lfstatus = False
     if modified or added or removed or deleted:
         raise util.Abort(_('outstanding uncommitted changes'))
     return orig(ui, repo, *pats, **opts)
 
 def overrideforget(orig, ui, repo, *pats, **opts):
     installnormalfilesmatchfn(repo[None].manifest())
     result = orig(ui, repo, *pats, **opts)
     restorematchfn()
     m = scmutil.match(repo[None], pats, opts)
 
     try:
         repo.lfstatus = True
         s = repo.status(match=m, clean=True)
     finally:
         repo.lfstatus = False
     forget = sorted(s[0] + s[1] + s[3] + s[6])
     forget = [f for f in forget if lfutil.standin(f) in repo[None].manifest()]
 
     for f in forget:
         if lfutil.standin(f) not in repo.dirstate and not \
                 os.path.isdir(m.rel(lfutil.standin(f))):
             ui.warn(_('not removing %s: file is already untracked\n')
                     % m.rel(f))
             result = 1
 
     for f in forget:
         if ui.verbose or not m.exact(f):
             ui.status(_('removing %s\n') % m.rel(f))
 
     # Need to lock because standin files are deleted then removed from the
     # repository and we could race in-between.
     wlock = repo.wlock()
     try:
         lfdirstate = lfutil.openlfdirstate(ui, repo)
         for f in forget:
             if lfdirstate[f] == 'a':
                 lfdirstate.drop(f)
             else:
                 lfdirstate.remove(f)
         lfdirstate.write()
         lfutil.reporemove(repo, [lfutil.standin(f) for f in forget],
             unlink=True)
     finally:
         wlock.release()
 
     return result
 
 def getoutgoinglfiles(ui, repo, dest=None, **opts):
     dest = ui.expandpath(dest or 'default-push', dest or 'default')
     dest, branches = hg.parseurl(dest, opts.get('branch'))
     revs, checkout = hg.addbranchrevs(repo, repo, branches, opts.get('rev'))
     if revs:
         revs = [repo.lookup(rev) for rev in scmutil.revrange(repo, revs)]
 
     try:
         remote = hg.peer(repo, opts, dest)
     except error.RepoError:
         return None
     o = lfutil.findoutgoing(repo, remote, False)
     if not o:
         return o
     o = repo.changelog.nodesbetween(o, revs)[0]
     if opts.get('newest_first'):
         o.reverse()
 
     toupload = set()
     for n in o:
         parents = [p for p in repo.changelog.parents(n) if p != node.nullid]
         ctx = repo[n]
         files = set(ctx.files())
         if len(parents) == 2:
             mc = ctx.manifest()
             mp1 = ctx.parents()[0].manifest()
             mp2 = ctx.parents()[1].manifest()
             for f in mp1:
                 if f not in mc:
                         files.add(f)
             for f in mp2:
                 if f not in mc:
                     files.add(f)
             for f in mc:
                 if mc[f] != mp1.get(f, None) or mc[f] != mp2.get(f, None):
                     files.add(f)
         toupload = toupload.union(
             set([f for f in files if lfutil.isstandin(f) and f in ctx]))
     return toupload
 
 def overrideoutgoing(orig, ui, repo, dest=None, **opts):
     result = orig(ui, repo, dest, **opts)
 
     if opts.pop('large', None):
         toupload = getoutgoinglfiles(ui, repo, dest, **opts)
         if toupload is None:
             ui.status(_('largefiles: No remote repo\n'))
         elif not toupload:
             ui.status(_('largefiles: no files to upload\n'))
         else:
             ui.status(_('largefiles to upload:\n'))
             for file in toupload:
                 ui.status(lfutil.splitstandin(file) + '\n')
             ui.status('\n')
 
     return result
 
 def overridesummary(orig, ui, repo, *pats, **opts):
     try:
         repo.lfstatus = True
         orig(ui, repo, *pats, **opts)
     finally:
         repo.lfstatus = False
 
     if opts.pop('large', None):
         toupload = getoutgoinglfiles(ui, repo, None, **opts)
         if toupload is None:
             ui.status(_('largefiles: No remote repo\n'))
         elif not toupload:
             ui.status(_('largefiles: (no files to upload)\n'))
         else:
             ui.status(_('largefiles: %d to upload\n') % len(toupload))
 
 def scmutiladdremove(orig, repo, pats=[], opts={}, dry_run=None,
                      similarity=None):
     if not lfutil.islfilesrepo(repo):
         return orig(repo, pats, opts, dry_run, similarity)
     # Get the list of missing largefiles so we can remove them
     lfdirstate = lfutil.openlfdirstate(repo.ui, repo)
     s = lfdirstate.status(match_.always(repo.root, repo.getcwd()), [], False,
         False, False)
     (unsure, modified, added, removed, missing, unknown, ignored, clean) = s
 
     # 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 missing:
         m = [repo.wjoin(f) for f in missing]
         repo._isaddremove = True
         removelargefiles(repo.ui, repo, *m, **opts)
         repo._isaddremove = False
     # Call into the normal add code, and any files that *should* be added as
     # largefiles will be
     addlargefiles(repo.ui, repo, *pats, **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 installing a matcher that will ignore them.
     installnormalfilesmatchfn(repo[None].manifest())
     result = orig(repo, pats, opts, dry_run, similarity)
     restorematchfn()
     return result
 
 # Calling purge with --all will cause the largefiles to be deleted.
 # Override repo.status to prevent this from happening.
 def overridepurge(orig, ui, repo, *dirs, **opts):
     oldstatus = repo.status
     def overridestatus(node1='.', 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)
         modified, added, removed, deleted, unknown, ignored, clean = r
         unknown = [f for f in unknown if lfdirstate[f] == '?']
         ignored = [f for f in ignored if lfdirstate[f] == '?']
         return modified, added, removed, deleted, unknown, ignored, clean
     repo.status = overridestatus
     orig(ui, repo, *dirs, **opts)
     repo.status = oldstatus
 
 def overriderollback(orig, ui, repo, **opts):
     result = orig(ui, repo, **opts)
     merge.update(repo, node=None, branchmerge=False, force=True,
         partial=lfutil.isstandin)
     wlock = repo.wlock()
     try:
         lfdirstate = lfutil.openlfdirstate(ui, repo)
         lfiles = lfutil.listlfiles(repo)
         oldlfiles = lfutil.listlfiles(repo, repo[None].parents()[0].rev())
         for file in lfiles:
             if file in oldlfiles:
                 lfdirstate.normallookup(file)
             else:
                 lfdirstate.add(file)
         lfdirstate.write()
     finally:
         wlock.release()
     return result
 
 def overridetransplant(orig, ui, repo, *revs, **opts):
     try:
         oldstandins = lfutil.getstandinsstate(repo)
         repo._istransplanting = True
         result = orig(ui, repo, *revs, **opts)
         newstandins = lfutil.getstandinsstate(repo)
         filelist = lfutil.getlfilestoupdate(oldstandins, newstandins)
         lfcommands.updatelfiles(repo.ui, repo, filelist=filelist,
                                 printmessage=True)
     finally:
         repo._istransplanting = False
     return result
 
 def overridecat(orig, ui, repo, file1, *pats, **opts):
     ctx = scmutil.revsingle(repo, opts.get('rev'))
     if not lfutil.standin(file1) in ctx:
         result = orig(ui, repo, file1, *pats, **opts)
         return result
     return lfcommands.catlfile(repo, file1, ctx.rev(), opts.get('output'))
+
+def mercurialsinkbefore(orig, sink):
+    sink.repo._isconverting = True
+    orig(sink)
+
+def mercurialsinkafter(orig, sink):
+    sink.repo._isconverting = False
+    orig(sink)

hgext/largefiles/uisetup.py

@@ -1,170 +1,177
 # 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.
 
 '''setup for largefiles extension: uisetup'''
 
 from mercurial import archival, cmdutil, commands, extensions, filemerge, hg, \
     httppeer, localrepo, merge, scmutil, sshpeer, sshserver, wireproto
 from mercurial.i18n import _
 from mercurial.hgweb import hgweb_mod, protocol, webcommands
 from mercurial.subrepo import hgsubrepo
 
 import overrides
 import proto
 
 def uisetup(ui):
     # Disable auto-status for some commands which assume that all
     # files in the result are under Mercurial's control
 
     entry = extensions.wrapcommand(commands.table, 'add',
                                    overrides.overrideadd)
     addopt = [('', 'large', None, _('add as largefile')),
               ('', 'normal', None, _('add as normal file')),
               ('', 'lfsize', '', _('add all files above this size '
                                    '(in megabytes) as largefiles '
                                    '(default: 10)'))]
     entry[1].extend(addopt)
 
     # The scmutil function is called both by the (trivial) addremove command,
     # and in the process of handling commit -A (issue3542)
     entry = extensions.wrapfunction(scmutil, 'addremove',
                                     overrides.scmutiladdremove)
     entry = extensions.wrapcommand(commands.table, 'remove',
                                    overrides.overrideremove)
     entry = extensions.wrapcommand(commands.table, 'forget',
                                    overrides.overrideforget)
 
     # Subrepos call status function
     entry = extensions.wrapcommand(commands.table, 'status',
                                    overrides.overridestatus)
     entry = extensions.wrapfunction(hgsubrepo, 'status',
                                     overrides.overridestatusfn)
 
     entry = extensions.wrapcommand(commands.table, 'log',
                                    overrides.overridelog)
     entry = extensions.wrapcommand(commands.table, 'rollback',
                                    overrides.overriderollback)
     entry = extensions.wrapcommand(commands.table, 'verify',
                                    overrides.overrideverify)
 
     verifyopt = [('', 'large', None, _('verify largefiles')),
                  ('', 'lfa', None,
                      _('verify all revisions of largefiles not just current')),
                  ('', 'lfc', None,
                      _('verify largefile contents not just existence'))]
     entry[1].extend(verifyopt)
 
     entry = extensions.wrapcommand(commands.table, 'outgoing',
         overrides.overrideoutgoing)
     outgoingopt = [('', 'large', None, _('display outgoing largefiles'))]
     entry[1].extend(outgoingopt)
     entry = extensions.wrapcommand(commands.table, 'summary',
                                    overrides.overridesummary)
     summaryopt = [('', 'large', None, _('display outgoing largefiles'))]
     entry[1].extend(summaryopt)
 
     entry = extensions.wrapcommand(commands.table, 'update',
                                    overrides.overrideupdate)
     entry = extensions.wrapcommand(commands.table, 'pull',
                                    overrides.overridepull)
     pullopt = [('', 'all-largefiles', None,
                  _('download all pulled versions of largefiles'))]
     entry[1].extend(pullopt)
     entry = extensions.wrapcommand(commands.table, 'clone',
                                    overrides.overrideclone)
     cloneopt = [('', 'all-largefiles', None,
                  _('download all versions of all largefiles'))]
     entry[1].extend(cloneopt)
     entry = extensions.wrapfunction(hg, 'clone', overrides.hgclone)
 
     entry = extensions.wrapcommand(commands.table, 'cat',
                                    overrides.overridecat)
     entry = extensions.wrapfunction(merge, '_checkunknownfile',
                                     overrides.overridecheckunknownfile)
     entry = extensions.wrapfunction(merge, 'manifestmerge',
                                     overrides.overridemanifestmerge)
     entry = extensions.wrapfunction(filemerge, 'filemerge',
                                     overrides.overridefilemerge)
     entry = extensions.wrapfunction(cmdutil, 'copy',
                                     overrides.overridecopy)
 
     # Summary calls dirty on the subrepos
     entry = extensions.wrapfunction(hgsubrepo, 'dirty',
                                     overrides.overridedirty)
 
     # Backout calls revert so we need to override both the command and the
     # function
     entry = extensions.wrapcommand(commands.table, 'revert',
                                    overrides.overriderevert)
     entry = extensions.wrapfunction(commands, 'revert',
                                     overrides.overriderevert)
 
     # clone uses hg._update instead of hg.update even though they are the
     # same function... so wrap both of them)
     extensions.wrapfunction(hg, 'update', overrides.hgupdate)
     extensions.wrapfunction(hg, '_update', overrides.hgupdate)
     extensions.wrapfunction(hg, 'clean', overrides.hgclean)
     extensions.wrapfunction(hg, 'merge', overrides.hgmerge)
 
     extensions.wrapfunction(archival, 'archive', overrides.overridearchive)
     extensions.wrapfunction(hgsubrepo, 'archive', overrides.hgsubrepoarchive)
     extensions.wrapfunction(cmdutil, 'bailifchanged',
                             overrides.overridebailifchanged)
 
     # create the new wireproto commands ...
     wireproto.commands['putlfile'] = (proto.putlfile, 'sha')
     wireproto.commands['getlfile'] = (proto.getlfile, 'sha')
     wireproto.commands['statlfile'] = (proto.statlfile, 'sha')
 
     # ... and wrap some existing ones
     wireproto.commands['capabilities'] = (proto.capabilities, '')
     wireproto.commands['heads'] = (proto.heads, '')
     wireproto.commands['lheads'] = (wireproto.heads, '')
 
     # make putlfile behave the same as push and {get,stat}lfile behave
     # the same as pull w.r.t. permissions checks
     hgweb_mod.perms['putlfile'] = 'push'
     hgweb_mod.perms['getlfile'] = 'pull'
     hgweb_mod.perms['statlfile'] = 'pull'
 
     extensions.wrapfunction(webcommands, 'decodepath', overrides.decodepath)
 
     # the hello wireproto command uses wireproto.capabilities, so it won't see
     # our largefiles capability unless we replace the actual function as well.
     proto.capabilitiesorig = wireproto.capabilities
     wireproto.capabilities = proto.capabilities
 
     # these let us reject non-largefiles clients and make them display
     # our error messages
     protocol.webproto.refuseclient = proto.webprotorefuseclient
     sshserver.sshserver.refuseclient = proto.sshprotorefuseclient
 
     # can't do this in reposetup because it needs to have happened before
     # wirerepo.__init__ is called
     proto.ssholdcallstream = sshpeer.sshpeer._callstream
     proto.httpoldcallstream = httppeer.httppeer._callstream
     sshpeer.sshpeer._callstream = proto.sshrepocallstream
     httppeer.httppeer._callstream = proto.httprepocallstream
 
     # don't die on seeing a repo with the largefiles requirement
     localrepo.localrepository.supported |= set(['largefiles'])
 
     # override some extensions' stuff as well
     for name, module in extensions.extensions():
         if name == 'fetch':
             extensions.wrapcommand(getattr(module, 'cmdtable'), 'fetch',
                 overrides.overridefetch)
         if name == 'purge':
             extensions.wrapcommand(getattr(module, 'cmdtable'), 'purge',
                 overrides.overridepurge)
         if name == 'rebase':
             extensions.wrapcommand(getattr(module, 'cmdtable'), 'rebase',
                 overrides.overriderebase)
         if name == 'transplant':
             extensions.wrapcommand(getattr(module, 'cmdtable'), 'transplant',
                 overrides.overridetransplant)
+        if name == 'convert':
+            convcmd = getattr(module, 'convcmd')
+            hgsink = getattr(convcmd, 'mercurial_sink')
+            extensions.wrapfunction(hgsink, 'before',
+                                    overrides.mercurialsinkbefore)
+            extensions.wrapfunction(hgsink, 'after',
+                                    overrides.mercurialsinkafter)

tests/test-lfconvert.t

@@ -1,338 +1,357
   $ USERCACHE="$TESTTMP/cache"; export USERCACHE
   $ mkdir "${USERCACHE}"
   $ cat >> $HGRCPATH <<EOF
   > [extensions]
   > largefiles =
   > share =
   > graphlog =
   > mq =
   > convert =
   > [largefiles]
   > minsize = 0.5
   > patterns = **.other
   >     **.dat
   > usercache=${USERCACHE}
   > EOF
 
 "lfconvert" works
   $ hg init bigfile-repo
   $ cd bigfile-repo
   $ cat >> .hg/hgrc <<EOF
   > [extensions]
   > largefiles = !
   > EOF
   $ mkdir sub
   $ dd if=/dev/zero bs=1k count=256 > large 2> /dev/null
   $ dd if=/dev/zero bs=1k count=256 > large2 2> /dev/null
   $ echo normal > normal1
   $ echo alsonormal > sub/normal2
   $ dd if=/dev/zero bs=1k count=10 > sub/maybelarge.dat 2> /dev/null
   $ hg addremove
   adding large
   adding large2
   adding normal1
   adding sub/maybelarge.dat
   adding sub/normal2
   $ hg commit -m"add large, normal1" large normal1
   $ hg commit -m"add sub/*" sub
 
 Test tag parsing
   $ cat >> .hgtags <<EOF
   > IncorrectlyFormattedTag!
   > invalidhash sometag
   > 0123456789abcdef anothertag
   > EOF
   $ hg add .hgtags
   $ hg commit -m"add large2" large2 .hgtags
 
 Test link+rename largefile codepath
   $ [ -d .hg/largefiles ] && echo fail || echo pass
   pass
   $ cd ..
   $ hg lfconvert --size 0.2 bigfile-repo largefiles-repo
   initializing destination largefiles-repo
   skipping incorrectly formatted tag IncorrectlyFormattedTag!
   skipping incorrectly formatted id invalidhash
   no mapping for id 0123456789abcdef
 #if symlink
   $ hg --cwd bigfile-repo rename large2 large3
   $ ln -sf large bigfile-repo/large3
   $ hg --cwd bigfile-repo commit -m"make large2 a symlink" large2 large3
   $ hg lfconvert --size 0.2 bigfile-repo largefiles-repo-symlink
   initializing destination largefiles-repo-symlink
   skipping incorrectly formatted tag IncorrectlyFormattedTag!
   skipping incorrectly formatted id invalidhash
   no mapping for id 0123456789abcdef
   abort: renamed/copied largefile large3 becomes symlink
   [255]
 #endif
   $ cd bigfile-repo
   $ hg strip --no-backup 2
   0 files updated, 0 files merged, 2 files removed, 0 files unresolved
   $ cd ..
   $ rm -rf largefiles-repo largefiles-repo-symlink
 
   $ hg lfconvert --size 0.2 bigfile-repo largefiles-repo
   initializing destination largefiles-repo
 
 "lfconvert" converts content correctly
   $ cd largefiles-repo
   $ hg up
   4 files updated, 0 files merged, 0 files removed, 0 files unresolved
   getting changed largefiles
   2 largefiles updated, 0 removed
   $ hg locate
   .hglf/large
   .hglf/sub/maybelarge.dat
   normal1
   sub/normal2
   $ cat normal1
   normal
   $ cat sub/normal2
   alsonormal
   $ "$TESTDIR/md5sum.py" large sub/maybelarge.dat
   ec87a838931d4d5d2e94a04644788a55  large
   1276481102f218c981e0324180bafd9f  sub/maybelarge.dat
 
 "lfconvert" adds 'largefiles' to .hg/requires.
   $ cat .hg/requires
   largefiles
   revlogv1
   fncache
   store
   dotencode
 
 "lfconvert" includes a newline at the end of the standin files.
   $ cat .hglf/large .hglf/sub/maybelarge.dat
   2e000fa7e85759c7f4c254d4d9c33ef481e459a7
   34e163be8e43c5631d8b92e9c43ab0bf0fa62b9c
   $ cd ..
 
 add some changesets to rename/remove/merge
   $ cd bigfile-repo
   $ hg mv -q sub stuff
   $ hg commit -m"rename sub/ to stuff/"
   $ hg update -q 1
   $ echo blah >> normal3
   $ echo blah >> sub/normal2
   $ echo blah >> sub/maybelarge.dat
   $ "$TESTDIR/md5sum.py" sub/maybelarge.dat
   1dd0b99ff80e19cff409702a1d3f5e15  sub/maybelarge.dat
   $ hg commit -A -m"add normal3, modify sub/*"
   adding normal3
   created new head
   $ hg rm large normal3
   $ hg commit -q -m"remove large, normal3"
   $ hg merge
   merging sub/maybelarge.dat and stuff/maybelarge.dat to stuff/maybelarge.dat
   warning: $TESTTMP/bigfile-repo/stuff/maybelarge.dat looks like a binary file. (glob)
   merging stuff/maybelarge.dat incomplete! (edit conflicts, then use 'hg resolve --mark')
   merging sub/normal2 and stuff/normal2 to stuff/normal2
   0 files updated, 1 files merged, 0 files removed, 1 files unresolved
   use 'hg resolve' to retry unresolved file merges or 'hg update -C .' to abandon
   [1]
   $ hg cat -r . sub/maybelarge.dat > stuff/maybelarge.dat
   $ hg resolve -m stuff/maybelarge.dat
   $ hg commit -m"merge"
   $ hg glog --template "{rev}:{node|short}  {desc|firstline}\n"
   @    5:4884f215abda  merge
   |\
   | o  4:7285f817b77e  remove large, normal3
   | |
   | o  3:67e3892e3534  add normal3, modify sub/*
   | |
   o |  2:c96c8beb5d56  rename sub/ to stuff/
   |/
   o  1:020c65d24e11  add sub/*
   |
   o  0:117b8328f97a  add large, normal1
   
   $ cd ..
 
 lfconvert with rename, merge, and remove
   $ rm -rf largefiles-repo
   $ hg lfconvert --size 0.2 bigfile-repo largefiles-repo
   initializing destination largefiles-repo
   $ cd largefiles-repo
   $ hg glog --template "{rev}:{node|short}  {desc|firstline}\n"
   o    5:8e05f5f2b77e  merge
   |\
   | o  4:a5a02de7a8e4  remove large, normal3
   | |
   | o  3:55759520c76f  add normal3, modify sub/*
   | |
   o |  2:261ad3f3f037  rename sub/ to stuff/
   |/
   o  1:334e5237836d  add sub/*
   |
   o  0:d4892ec57ce2  add large, normal1
   
   $ hg locate -r 2
   .hglf/large
   .hglf/stuff/maybelarge.dat
   normal1
   stuff/normal2
   $ hg locate -r 3
   .hglf/large
   .hglf/sub/maybelarge.dat
   normal1
   normal3
   sub/normal2
   $ hg locate -r 4
   .hglf/sub/maybelarge.dat
   normal1
   sub/normal2
   $ hg locate -r 5
   .hglf/stuff/maybelarge.dat
   normal1
   stuff/normal2
   $ hg update
   3 files updated, 0 files merged, 0 files removed, 0 files unresolved
   getting changed largefiles
   1 largefiles updated, 0 removed
   $ cat stuff/normal2
   alsonormal
   blah
   $ "$TESTDIR/md5sum.py" stuff/maybelarge.dat
   1dd0b99ff80e19cff409702a1d3f5e15  stuff/maybelarge.dat
   $ cat .hglf/stuff/maybelarge.dat
   76236b6a2c6102826c61af4297dd738fb3b1de38
   $ cd ..
 
 "lfconvert" error cases
   $ hg lfconvert http://localhost/foo foo
   abort: http://localhost/foo is not a local Mercurial repo
   [255]
   $ hg lfconvert foo ssh://localhost/foo
   abort: ssh://localhost/foo is not a local Mercurial repo
   [255]
   $ hg lfconvert nosuchrepo foo
   abort: repository nosuchrepo not found!
   [255]
   $ hg share -q -U bigfile-repo shared
   $ printf 'bogus' > shared/.hg/sharedpath
   $ hg lfconvert shared foo
   abort: .hg/sharedpath points to nonexistent directory $TESTTMP/bogus! (glob)
   [255]
   $ hg lfconvert bigfile-repo largefiles-repo
   initializing destination largefiles-repo
   abort: repository largefiles-repo already exists!
   [255]
 
 add another largefile to the new largefiles repo
   $ cd largefiles-repo
   $ dd if=/dev/zero bs=1k count=1k > anotherlarge 2> /dev/null
   $ hg add --lfsize=1 anotherlarge
   $ hg commit -m "add anotherlarge (should be a largefile)"
   $ cat .hglf/anotherlarge
   3b71f43ff30f4b15b5cd85dd9e95ebc7e84eb5a3
   $ cd ..
 
 round-trip: converting back to a normal (non-largefiles) repo with
 "lfconvert --to-normal" should give the same as ../bigfile-repo
   $ cd largefiles-repo
   $ hg lfconvert --to-normal . ../normal-repo
   initializing destination ../normal-repo
   $ cd ../normal-repo
   $ cat >> .hg/hgrc <<EOF
   > [extensions]
   > largefiles = !
   > EOF
 
 # Hmmm: the changeset ID for rev 5 is different from the original
 # normal repo (../bigfile-repo), because the changelog filelist
 # differs between the two incarnations of rev 5: this repo includes
 # 'large' in the list, but ../bigfile-repo does not. Since rev 5
 # removes 'large' relative to the first parent in both repos, it seems
 # to me that lfconvert is doing a *better* job than
 # "hg remove" + "hg merge" + "hg commit".
 #  $ hg -R ../bigfile-repo debugdata -c 5
 #  $ hg debugdata -c 5
   $ hg glog --template "{rev}:{node|short}  {desc|firstline}\n"
   o  6:1635824e6f59  add anotherlarge (should be a largefile)
   |
   o    5:7215f8deeaaf  merge
   |\
   | o  4:7285f817b77e  remove large, normal3
   | |
   | o  3:67e3892e3534  add normal3, modify sub/*
   | |
   o |  2:c96c8beb5d56  rename sub/ to stuff/
   |/
   o  1:020c65d24e11  add sub/*
   |
   o  0:117b8328f97a  add large, normal1
   
   $ hg update
   4 files updated, 0 files merged, 0 files removed, 0 files unresolved
   $ hg locate
   anotherlarge
   normal1
   stuff/maybelarge.dat
   stuff/normal2
   $ [ -d .hg/largefiles ] && echo fail || echo pass
   pass
 
   $ cd ..
 
+Clearing the usercache ensures that commitctx doesn't try to cache largefiles
+from the working dir on a convert.
+  $ rm "${USERCACHE}"/*
   $ hg convert largefiles-repo
   assuming destination largefiles-repo-hg
   initializing destination largefiles-repo-hg repository
   scanning source...
   sorting...
   converting...
   6 add large, normal1
   5 add sub/*
   4 rename sub/ to stuff/
   3 add normal3, modify sub/*
   2 remove large, normal3
   1 merge
   0 add anotherlarge (should be a largefile)
 
   $ hg -R largefiles-repo-hg glog --template "{rev}:{node|short}  {desc|firstline}\n"
   o  6:17126745edfd  add anotherlarge (should be a largefile)
   |
   o    5:9cc5aa7204f0  merge
   |\
   | o  4:a5a02de7a8e4  remove large, normal3
   | |
   | o  3:55759520c76f  add normal3, modify sub/*
   | |
   o |  2:261ad3f3f037  rename sub/ to stuff/
   |/
   o  1:334e5237836d  add sub/*
   |
   o  0:d4892ec57ce2  add large, normal1
   
+Verify will fail (for now) if the usercache is purged before converting, since
+largefiles are not cached in the converted repo's local store by the conversion
+process.
   $ hg -R largefiles-repo-hg verify --large --lfa
   checking changesets
   checking manifests
   crosschecking files in changesets and manifests
   checking files
   8 files, 7 changesets, 12 total revisions
   searching 7 changesets for largefiles
+  changeset 0:d4892ec57ce2: large missing
+    (looked for hash 2e000fa7e85759c7f4c254d4d9c33ef481e459a7)
+  changeset 1:334e5237836d: sub/maybelarge.dat missing
+    (looked for hash 34e163be8e43c5631d8b92e9c43ab0bf0fa62b9c)
+  changeset 2:261ad3f3f037: stuff/maybelarge.dat missing
+    (looked for hash 34e163be8e43c5631d8b92e9c43ab0bf0fa62b9c)
+  changeset 3:55759520c76f: sub/maybelarge.dat missing
+    (looked for hash 76236b6a2c6102826c61af4297dd738fb3b1de38)
+  changeset 5:9cc5aa7204f0: stuff/maybelarge.dat missing
+    (looked for hash 76236b6a2c6102826c61af4297dd738fb3b1de38)
+  changeset 6:17126745edfd: anotherlarge missing
+    (looked for hash 3b71f43ff30f4b15b5cd85dd9e95ebc7e84eb5a3)
   verified existence of 6 revisions of 4 largefiles
+  [1]
   $ hg -R largefiles-repo-hg showconfig paths
 
 
 Avoid a traceback if a largefile isn't available (issue3519)
 
 Ensure the largefile can be cached in the source if necessary
   $ hg clone -U largefiles-repo issue3519
-  $ rm "${USERCACHE}"/*
+  $ rm -f "${USERCACHE}"/*
   $ hg lfconvert --to-normal issue3519 normalized3519
   initializing destination normalized3519
 
 Ensure the abort message is useful if a largefile is entirely unavailable
   $ rm -rf normalized3519
   $ rm "${USERCACHE}"/*
   $ rm issue3519/.hg/largefiles/*
   $ rm largefiles-repo/.hg/largefiles/*
   $ hg lfconvert --to-normal issue3519 normalized3519
   initializing destination normalized3519
   large: can't get file locally
   (no default or default-push path set in hgrc)
   abort: missing largefile 'large' from revision d4892ec57ce212905215fad1d9018f56b99202ad
   [255]