largefiles: refactor downloading of all largefiles to generic function

hgext/largefiles/lfcommands.py

@@ -1,517 +1,535 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.
 
 '''High-level command function for lfconvert, plus the cmdtable.'''
 
 import os
 import shutil
 
-from mercurial import util, match as match_, hg, node, context, error, cmdutil
+from mercurial import util, match as match_, hg, node, context, error, \
+    cmdutil, scmutil
 from mercurial.i18n import _
 
 import lfutil
 import basestore
 
 # -- Commands ----------------------------------------------------------
 
 def lfconvert(ui, src, dest, *pats, **opts):
     '''convert a normal repository to a largefiles repository
 
     Convert repository SOURCE to a new repository DEST, identical to
     SOURCE except that certain files will be converted as largefiles:
     specifically, any file that matches any PATTERN *or* whose size is
     above the minimum size threshold is converted as a largefile. The
     size used to determine whether or not to track a file as a
     largefile is the size of the first version of the file. The
     minimum size can be specified either with --size or in
     configuration as ``largefiles.size``.
 
     After running this command you will need to make sure that
     largefiles is enabled anywhere you intend to push the new
     repository.
 
     Use --to-normal to convert largefiles back to normal files; after
     this, the DEST repository can be used without largefiles at all.'''
 
     if opts['to_normal']:
         tolfile = False
     else:
         tolfile = True
         size = lfutil.getminsize(ui, True, opts.get('size'), default=None)
 
     if not hg.islocal(src):
         raise util.Abort(_('%s is not a local Mercurial repo') % src)
     if not hg.islocal(dest):
         raise util.Abort(_('%s is not a local Mercurial repo') % dest)
 
     rsrc = hg.repository(ui, src)
     ui.status(_('initializing destination %s\n') % dest)
     rdst = hg.repository(ui, dest, create=True)
 
     success = False
     try:
         # Lock destination to prevent modification while it is converted to.
         # Don't need to lock src because we are just reading from its history
         # which can't change.
         dstlock = rdst.lock()
 
         # Get a list of all changesets in the source.  The easy way to do this
         # is to simply walk the changelog, using changelog.nodesbewteen().
         # Take a look at mercurial/revlog.py:639 for more details.
         # Use a generator instead of a list to decrease memory usage
         ctxs = (rsrc[ctx] for ctx in rsrc.changelog.nodesbetween(None,
             rsrc.heads())[0])
         revmap = {node.nullid: node.nullid}
         if tolfile:
             lfiles = set()
             normalfiles = set()
             if not pats:
                 pats = ui.configlist(lfutil.longname, 'patterns', default=[])
             if pats:
                 matcher = match_.match(rsrc.root, '', list(pats))
             else:
                 matcher = None
 
             lfiletohash = {}
             for ctx in ctxs:
                 ui.progress(_('converting revisions'), ctx.rev(),
                     unit=_('revision'), total=rsrc['tip'].rev())
                 _lfconvert_addchangeset(rsrc, rdst, ctx, revmap,
                     lfiles, normalfiles, matcher, size, lfiletohash)
             ui.progress(_('converting revisions'), None)
 
             if os.path.exists(rdst.wjoin(lfutil.shortname)):
                 shutil.rmtree(rdst.wjoin(lfutil.shortname))
 
             for f in lfiletohash.keys():
                 if os.path.isfile(rdst.wjoin(f)):
                     os.unlink(rdst.wjoin(f))
                 try:
                     os.removedirs(os.path.dirname(rdst.wjoin(f)))
                 except OSError:
                     pass
 
             # If there were any files converted to largefiles, add largefiles
             # to the destination repository's requirements.
             if lfiles:
                 rdst.requirements.add('largefiles')
                 rdst._writerequirements()
         else:
             for ctx in ctxs:
                 ui.progress(_('converting revisions'), ctx.rev(),
                     unit=_('revision'), total=rsrc['tip'].rev())
                 _addchangeset(ui, rsrc, rdst, ctx, revmap)
 
             ui.progress(_('converting revisions'), None)
         success = True
     finally:
         if not success:
             # we failed, remove the new directory
             shutil.rmtree(rdst.root)
         dstlock.release()
 
 def _addchangeset(ui, rsrc, rdst, ctx, revmap):
  # Convert src parents to dst parents
     parents = _convertparents(ctx, revmap)
 
     # Generate list of changed files
     files = _getchangedfiles(ctx, parents)
 
     def getfilectx(repo, memctx, f):
         if lfutil.standin(f) in files:
             # if the file isn't in the manifest then it was removed
             # or renamed, raise IOError to indicate this
             try:
                 fctx = ctx.filectx(lfutil.standin(f))
             except error.LookupError:
                 raise IOError
             renamed = fctx.renamed()
             if renamed:
                 renamed = lfutil.splitstandin(renamed[0])
 
             hash = fctx.data().strip()
             path = lfutil.findfile(rsrc, hash)
             ### TODO: What if the file is not cached?
             data = ''
             fd = None
             try:
                 fd = open(path, 'rb')
                 data = fd.read()
             finally:
                 if fd:
                     fd.close()
             return context.memfilectx(f, data, 'l' in fctx.flags(),
                                       'x' in fctx.flags(), renamed)
         else:
             return _getnormalcontext(repo.ui, ctx, f, revmap)
 
     dstfiles = []
     for file in files:
         if lfutil.isstandin(file):
             dstfiles.append(lfutil.splitstandin(file))
         else:
             dstfiles.append(file)
     # Commit
     _commitcontext(rdst, parents, ctx, dstfiles, getfilectx, revmap)
 
 def _lfconvert_addchangeset(rsrc, rdst, ctx, revmap, lfiles, normalfiles,
         matcher, size, lfiletohash):
     # Convert src parents to dst parents
     parents = _convertparents(ctx, revmap)
 
     # Generate list of changed files
     files = _getchangedfiles(ctx, parents)
 
     dstfiles = []
     for f in files:
         if f not in lfiles and f not in normalfiles:
             islfile = _islfile(f, ctx, matcher, size)
             # If this file was renamed or copied then copy
             # the lfileness of its predecessor
             if f in ctx.manifest():
                 fctx = ctx.filectx(f)
                 renamed = fctx.renamed()
                 renamedlfile = renamed and renamed[0] in lfiles
                 islfile |= renamedlfile
                 if 'l' in fctx.flags():
                     if renamedlfile:
                         raise util.Abort(
                             _('renamed/copied largefile %s becomes symlink')
                             % f)
                     islfile = False
             if islfile:
                 lfiles.add(f)
             else:
                 normalfiles.add(f)
 
         if f in lfiles:
             dstfiles.append(lfutil.standin(f))
             # largefile in manifest if it has not been removed/renamed
             if f in ctx.manifest():
                 fctx = ctx.filectx(f)
                 if 'l' in fctx.flags():
                     renamed = fctx.renamed()
                     if renamed and renamed[0] in lfiles:
                         raise util.Abort(_('largefile %s becomes symlink') % f)
 
                 # largefile was modified, update standins
                 fullpath = rdst.wjoin(f)
                 util.makedirs(os.path.dirname(fullpath))
                 m = util.sha1('')
                 m.update(ctx[f].data())
                 hash = m.hexdigest()
                 if f not in lfiletohash or lfiletohash[f] != hash:
                     try:
                         fd = open(fullpath, 'wb')
                         fd.write(ctx[f].data())
                     finally:
                         if fd:
                             fd.close()
                     executable = 'x' in ctx[f].flags()
                     os.chmod(fullpath, lfutil.getmode(executable))
                     lfutil.writestandin(rdst, lfutil.standin(f), hash,
                         executable)
                     lfiletohash[f] = hash
         else:
             # normal file
             dstfiles.append(f)
 
     def getfilectx(repo, memctx, f):
         if lfutil.isstandin(f):
             # if the file isn't in the manifest then it was removed
             # or renamed, raise IOError to indicate this
             srcfname = lfutil.splitstandin(f)
             try:
                 fctx = ctx.filectx(srcfname)
             except error.LookupError:
                 raise IOError
             renamed = fctx.renamed()
             if renamed:
                 # standin is always a largefile because largefile-ness
                 # doesn't change after rename or copy
                 renamed = lfutil.standin(renamed[0])
 
             return context.memfilectx(f, lfiletohash[srcfname] + '\n', 'l' in
                 fctx.flags(), 'x' in fctx.flags(), renamed)
         else:
             return _getnormalcontext(repo.ui, ctx, f, revmap)
 
     # Commit
     _commitcontext(rdst, parents, ctx, dstfiles, getfilectx, revmap)
 
 def _commitcontext(rdst, parents, ctx, dstfiles, getfilectx, revmap):
     mctx = context.memctx(rdst, parents, ctx.description(), dstfiles,
                           getfilectx, ctx.user(), ctx.date(), ctx.extra())
     ret = rdst.commitctx(mctx)
     rdst.setparents(ret)
     revmap[ctx.node()] = rdst.changelog.tip()
 
 # Generate list of changed files
 def _getchangedfiles(ctx, parents):
     files = set(ctx.files())
     if node.nullid not in parents:
         mc = ctx.manifest()
         mp1 = ctx.parents()[0].manifest()
         mp2 = ctx.parents()[1].manifest()
         files |= (set(mp1) | set(mp2)) - set(mc)
         for f in mc:
             if mc[f] != mp1.get(f, None) or mc[f] != mp2.get(f, None):
                 files.add(f)
     return files
 
 # Convert src parents to dst parents
 def _convertparents(ctx, revmap):
     parents = []
     for p in ctx.parents():
         parents.append(revmap[p.node()])
     while len(parents) < 2:
         parents.append(node.nullid)
     return parents
 
 # Get memfilectx for a normal file
 def _getnormalcontext(ui, ctx, f, revmap):
     try:
         fctx = ctx.filectx(f)
     except error.LookupError:
         raise IOError
     renamed = fctx.renamed()
     if renamed:
         renamed = renamed[0]
 
     data = fctx.data()
     if f == '.hgtags':
         data = _converttags (ui, revmap, data)
     return context.memfilectx(f, data, 'l' in fctx.flags(),
                               'x' in fctx.flags(), renamed)
 
 # Remap tag data using a revision map
 def _converttags(ui, revmap, data):
     newdata = []
     for line in data.splitlines():
         try:
             id, name = line.split(' ', 1)
         except ValueError:
             ui.warn(_('skipping incorrectly formatted tag %s\n'
                 % line))
             continue
         try:
             newid = node.bin(id)
         except TypeError:
             ui.warn(_('skipping incorrectly formatted id %s\n'
                 % id))
             continue
         try:
             newdata.append('%s %s\n' % (node.hex(revmap[newid]),
                 name))
         except KeyError:
             ui.warn(_('no mapping for id %s\n') % id)
             continue
     return ''.join(newdata)
 
 def _islfile(file, ctx, matcher, size):
     '''Return true if file should be considered a largefile, i.e.
     matcher matches it or it is larger than size.'''
     # never store special .hg* files as largefiles
     if file == '.hgtags' or file == '.hgignore' or file == '.hgsigs':
         return False
     if matcher and matcher(file):
         return True
     try:
         return ctx.filectx(file).size() >= size * 1024 * 1024
     except error.LookupError:
         return False
 
 def uploadlfiles(ui, rsrc, rdst, files):
     '''upload largefiles to the central store'''
 
     if not files:
         return
 
     store = basestore._openstore(rsrc, rdst, put=True)
 
     at = 0
     files = filter(lambda h: not store.exists(h), files)
     for hash in files:
         ui.progress(_('uploading largefiles'), at, unit='largefile',
                     total=len(files))
         source = lfutil.findfile(rsrc, hash)
         if not source:
             raise util.Abort(_('largefile %s missing from store'
                                ' (needs to be uploaded)') % hash)
         # XXX check for errors here
         store.put(source, hash)
         at += 1
     ui.progress(_('uploading largefiles'), None)
 
 def verifylfiles(ui, repo, all=False, contents=False):
     '''Verify that every big file revision in the current changeset
     exists in the central store.  With --contents, also verify that
     the contents of each big file revision are correct (SHA-1 hash
     matches the revision ID).  With --all, check every changeset in
     this repository.'''
     if all:
         # Pass a list to the function rather than an iterator because we know a
         # list will work.
         revs = range(len(repo))
     else:
         revs = ['.']
 
     store = basestore._openstore(repo)
     return store.verify(revs, contents=contents)
 
 def cachelfiles(ui, repo, node):
     '''cachelfiles ensures that all largefiles needed by the specified revision
     are present in the repository's largefile cache.
 
     returns a tuple (cached, missing).  cached is the list of files downloaded
     by this operation; missing is the list of files that were needed but could
     not be found.'''
     lfiles = lfutil.listlfiles(repo, node)
     toget = []
 
     for lfile in lfiles:
         # If we are mid-merge, then we have to trust the standin that is in the
         # working copy to have the correct hashvalue.  This is because the
         # original hg.merge() already updated the standin as part of the normal
         # merge process -- we just have to udpate the largefile to match.
         if (getattr(repo, "_ismerging", False) and
              os.path.exists(repo.wjoin(lfutil.standin(lfile)))):
             expectedhash = lfutil.readstandin(repo, lfile)
         else:
             expectedhash = repo[node][lfutil.standin(lfile)].data().strip()
 
         # if it exists and its hash matches, it might have been locally
         # modified before updating and the user chose 'local'.  in this case,
         # it will not be in any store, so don't look for it.
         if ((not os.path.exists(repo.wjoin(lfile)) or
              expectedhash != lfutil.hashfile(repo.wjoin(lfile))) and
             not lfutil.findfile(repo, expectedhash)):
             toget.append((lfile, expectedhash))
 
     if toget:
         store = basestore._openstore(repo)
         ret = store.get(toget)
         return ret
 
     return ([], [])
 
+def downloadlfiles(ui, repo, rev=None):
+    matchfn = scmutil.match(repo[None],
+                            [repo.wjoin(lfutil.shortname)], {})
+    def prepare(ctx, fns):
+        pass
+    totalsuccess = 0
+    totalmissing = 0
+    for ctx in cmdutil.walkchangerevs(repo, matchfn, {'rev' : rev},
+                                      prepare):
+        success, missing = cachelfiles(ui, repo, ctx.node())
+        totalsuccess += len(success)
+        totalmissing += len(missing)
+    ui.status(_("%d additional largefiles cached\n") % totalsuccess)
+    if totalmissing > 0:
+        ui.status(_("%d largefiles failed to download\n") % totalmissing)
+    return totalsuccess, totalmissing
+
 def updatelfiles(ui, repo, filelist=None, printmessage=True):
     wlock = repo.wlock()
     try:
         lfdirstate = lfutil.openlfdirstate(ui, repo)
         lfiles = set(lfutil.listlfiles(repo)) | set(lfdirstate)
 
         if filelist is not None:
             lfiles = [f for f in lfiles if f in filelist]
 
         printed = False
         if printmessage and lfiles:
             ui.status(_('getting changed largefiles\n'))
             printed = True
             cachelfiles(ui, repo, '.')
 
         updated, removed = 0, 0
         for i in map(lambda f: _updatelfile(repo, lfdirstate, f), lfiles):
             # increment the appropriate counter according to _updatelfile's
             # return value
             updated += i > 0 and i or 0
             removed -= i < 0 and i or 0
             if printmessage and (removed or updated) and not printed:
                 ui.status(_('getting changed largefiles\n'))
                 printed = True
 
         lfdirstate.write()
         if printed and printmessage:
             ui.status(_('%d largefiles updated, %d removed\n') % (updated,
                 removed))
     finally:
         wlock.release()
 
 def _updatelfile(repo, lfdirstate, lfile):
     '''updates a single largefile and copies the state of its standin from
     the repository's dirstate to its state in the lfdirstate.
 
     returns 1 if the file was modified, -1 if the file was removed, 0 if the
     file was unchanged, and None if the needed largefile was missing from the
     cache.'''
     ret = 0
     abslfile = repo.wjoin(lfile)
     absstandin = repo.wjoin(lfutil.standin(lfile))
     if os.path.exists(absstandin):
         if os.path.exists(absstandin+'.orig'):
             shutil.copyfile(abslfile, abslfile+'.orig')
         expecthash = lfutil.readstandin(repo, lfile)
         if (expecthash != '' and
             (not os.path.exists(abslfile) or
              expecthash != lfutil.hashfile(abslfile))):
             if not lfutil.copyfromcache(repo, expecthash, lfile):
                 # use normallookup() to allocate entry in largefiles dirstate,
                 # because lack of it misleads lfilesrepo.status() into
                 # recognition that such cache missing files are REMOVED.
                 lfdirstate.normallookup(lfile)
                 return None # don't try to set the mode
             ret = 1
         mode = os.stat(absstandin).st_mode
         if mode != os.stat(abslfile).st_mode:
             os.chmod(abslfile, mode)
             ret = 1
     else:
         # Remove lfiles for which the standin is deleted, unless the
         # lfile is added to the repository again. This happens when a
         # largefile is converted back to a normal file: the standin
         # disappears, but a new (normal) file appears as the lfile.
         if os.path.exists(abslfile) and lfile not in repo[None]:
             util.unlinkpath(abslfile)
             ret = -1
     state = repo.dirstate[lfutil.standin(lfile)]
     if state == 'n':
         # When rebasing, we need to synchronize the standin and the largefile,
         # because otherwise the largefile will get reverted.  But for commit's
         # sake, we have to mark the file as unclean.
         if getattr(repo, "_isrebasing", False):
            lfdirstate.normallookup(lfile)
         else:
             lfdirstate.normal(lfile)
     elif state == 'r':
         lfdirstate.remove(lfile)
     elif state == 'a':
         lfdirstate.add(lfile)
     elif state == '?':
         lfdirstate.drop(lfile)
     return ret
 
 def catlfile(repo, lfile, rev, filename):
     hash = lfutil.readstandin(repo, lfile, rev)
     if not lfutil.inusercache(repo.ui, hash):
         store = basestore._openstore(repo)
         success, missing = store.get([(lfile, hash)])
         if len(success) != 1:
             raise util.Abort(
                 _('largefile %s is not in cache and could not be downloaded')
                     % lfile)
     path = lfutil.usercachepath(repo.ui, hash)
     fpout = cmdutil.makefileobj(repo, filename)
     fpin = open(path, "rb")
     fpout.write(fpin.read())
     fpout.close()
     fpin.close()
     return 0
 
 # -- hg commands declarations ------------------------------------------------
 
 cmdtable = {
     'lfconvert': (lfconvert,
                   [('s', 'size', '',
                     _('minimum size (MB) for files to be converted '
                       'as largefiles'),
                     'SIZE'),
                   ('', 'to-normal', False,
                    _('convert from a largefiles repo to a normal repo')),
                   ],
                   _('hg lfconvert SOURCE DEST [FILE ...]')),
     }

hgext/largefiles/overrides.py

@@ -1,1065 +1,1053 b''
 # Copyright 2009-2010 Gregory P. Ward
 # Copyright 2009-2010 Intelerad Medical Systems Incorporated
 # Copyright 2010-2011 Fog Creek Software
 # Copyright 2010-2011 Unity Technologies
 #
 # This software may be used and distributed according to the terms of the
 # GNU General Public License version 2 or any later version.
 
 '''Overridden Mercurial commands and functions for the largefiles extension'''
 
 import 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:
             abovemin = (lfsize and
                         os.lstat(repo.wjoin(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(_('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))
 
     if after:
         remove, forget = deleted, []
         warn(modified + added + clean, _('file still exists'))
     else:
         remove, forget = deleted + clean, []
         warn(modified, _('file is modified'))
         warn(added, _('file has been marked for add'))
 
     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 inbetween.
     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)
     finally:
         wlock.release()
 
 # 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())
     orig(ui, repo, *pats, **opts)
     restorematchfn()
     removelargefiles(ui, repo, *pats, **opts)
 
 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
         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 overriden 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(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(srclfile, 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 = 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):
     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)
             revsprepull = len(repo)
             origpostincoming = commands.postincoming
             def _dummy(*args, **kwargs):
                 pass
             commands.postincoming = _dummy
             repo.lfpullsource = source
             if not source:
                 source = 'default'
             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:
         repo.lfpullsource = source
         if not source:
             source = 'default'
         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 direclty 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)
     return result
 
 def overrideclone(orig, ui, source, dest=None, **opts):
     result = hg.clone(ui, opts, source, dest,
                       pull=opts.get('pull'),
                       stream=opts.get('uncompressed'),
                       rev=opts.get('rev'),
                       update=True, # required for successful walkchangerevs
                       branch=opts.get('branch'))
     if result is None:
         return True
-    totalsuccess = 0
-    totalmissing = 0
     if opts.get('all_largefiles'):
         sourcerepo, destrepo = result
-        matchfn = scmutil.match(destrepo[None],
-                                [destrepo.wjoin(lfutil.shortname)], {})
-        def prepare(ctx, fns):
-            pass
-        for ctx in cmdutil.walkchangerevs(destrepo, matchfn, {'rev' : None},
-                                          prepare):
-            success, missing = lfcommands.cachelfiles(ui, destrepo, ctx.node())
-            totalsuccess += len(success)
-            totalmissing += len(missing)
-        ui.status(_("%d additional largefiles cached\n") % totalsuccess)
-        if totalmissing > 0:
-            ui.status(_("%d largefiles failed to download\n") % totalmissing)
-    return totalmissing != 0
+        success, missing = lfcommands.downloadlfiles(ui, destrepo, None)
+        return missing != 0
+    return result is None
 
 def overriderebase(orig, ui, repo, **opts):
     repo._isrebasing = True
     try:
         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)
             sub.archive(repo.ui, archiver, prefix)
 
     archiver.done()
 
 def hgsubrepoarchive(orig, repo, ui, archiver, prefix):
     rev = repo._state[1]
     ctx = repo._repo[rev]
 
     lfcommands.cachelfiles(ui, repo._repo, ctx.node())
 
     def write(name, mode, islink, getdata):
         if lfutil.isstandin(name):
             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)
         sub.archive(repo.ui, archiver, prefix)
 
 # 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())
     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))
 
     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 inbetween.
     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()
 
 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 revs]
 
     remoteui = hg.remoteui
 
     try:
         remote = hg.repository(remoteui(repo, opts), dest)
     except error.RepoError:
         return None
     o = lfutil.findoutgoing(repo, remote, False)
     if not o:
         return None
     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):
     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'))
         else:
             ui.status(_('largefiles to upload:\n'))
             for file in toupload:
                 ui.status(lfutil.splitstandin(file) + '\n')
             ui.status('\n')
 
 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'))
         else:
             ui.status(_('largefiles: %d to upload\n') % len(toupload))
 
 def overrideaddremove(orig, ui, repo, *pats, **opts):
     if not lfutil.islfilesrepo(repo):
         return orig(ui, repo, *pats, **opts)
     # Get the list of missing largefiles so we can remove them
     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
 
     # 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:
         repo._isaddremove = True
         removelargefiles(ui, repo, *missing, **opts)
         repo._isaddremove = False
     # Call into the normal add code, and any files that *should* be added as
     # largefiles will be
     addlargefiles(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(ui, repo, *pats, **opts)
     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):
     rev = opts.get('rev')
     if not lfutil.standin(file1) in repo[rev]:
         result = orig(ui, repo, file1, *pats, **opts)
         return result
     return lfcommands.catlfile(repo, file1, opts.get('rev'), opts.get('output'))