largefiles: factor out lfutil.getminsize()

hgext/largefiles/lfcommands.py

@@ -1,483 +1,473 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 functions: lfadd() et. al, plus the cmdtable.'''
 
 import os
 import shutil
 
 from mercurial import util, match as match_, hg, node, context, error
 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 source repository creating an identical repository, except that all
     files that match the patterns given, or are over the given size will be
     added as largefiles. 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. After
     running this command you will need to make sure that largefiles is enabled
     anywhere you intend to push the new repository.'''
 
     if opts['tonormal']:
         tolfile = False
     else:
         tolfile = True
-        size = opts['size']
-        if not size:
-            size = ui.config(lfutil.longname, 'size', default=None)
-            try:
-                size = int(size)
-            except ValueError:
-                raise util.Abort(_('largefiles.size must be integer, was %s\n')
-                                 % size)
-            except TypeError:
-                raise util.Abort(_('size must be specified'))
-
+        size = lfutil.getminsize(ui, True, opts.get('size'), default=None)
     try:
         rsrc = hg.repository(ui, src)
         if not rsrc.local():
             raise util.Abort(_('%s is not a local Mercurial repo') % src)
     except error.RepoError, err:
         ui.traceback()
         raise util.Abort(err.args[0])
     if os.path.exists(dest):
         if not os.path.isdir(dest):
             raise util.Abort(_('destination %s already exists') % dest)
         elif os.listdir(dest):
             raise util.Abort(_('destination %s is not empty') % dest)
     try:
         ui.status(_('initializing destination %s\n') % dest)
         rdst = hg.repository(ui, dest, create=True)
         if not rdst.local():
             raise util.Abort(_('%s is not a local Mercurial repo') % dest)
     except error.RepoError:
         ui.traceback()
         raise util.Abort(_('%s is not a repo') % dest)
 
     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.
         dst_lock = 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.config(lfutil.longname, 'patterns', default=())
                 if pats:
                     pats = pats.split(' ')
             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
 
         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)
         dst_lock.release()
 
 def _addchangeset(ui, rsrc, rdst, ctx, revmap):
  # Convert src parents to dst parents
     parents = []
     for p in ctx.parents():
         parents.append(revmap[p.node()])
     while len(parents) < 2:
         parents.append(node.nullid)
 
     # Generate list of changed files
     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)
 
     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:
             try:
                 fctx = ctx.filectx(f)
             except error.LookupError:
                 raise IOError()
             renamed = fctx.renamed()
             if renamed:
                 renamed = renamed[0]
             data = fctx.data()
             if f == '.hgtags':
                 newdata = []
                 for line in data.splitlines():
                     id, name = line.split(' ', 1)
                     newdata.append('%s %s\n' % (node.hex(revmap[node.bin(id)]),
                         name))
                 data = ''.join(newdata)
             return context.memfilectx(f, data, 'l' in fctx.flags(),
                                       'x' in fctx.flags(), renamed)
 
     dstfiles = []
     for file in files:
         if lfutil.isstandin(file):
             dstfiles.append(lfutil.splitstandin(file))
         else:
             dstfiles.append(file)
     # Commit
     mctx = context.memctx(rdst, parents, ctx.description(), dstfiles,
                           getfilectx, ctx.user(), ctx.date(), ctx.extra())
     ret = rdst.commitctx(mctx)
     rdst.dirstate.setparents(ret)
     revmap[ctx.node()] = rdst.changelog.tip()
 
 def _lfconvert_addchangeset(rsrc, rdst, ctx, revmap, lfiles, normalfiles,
         matcher, size, lfiletohash):
     # Convert src parents to dst parents
     parents = []
     for p in ctx.parents():
         parents.append(revmap[p.node()])
     while len(parents) < 2:
         parents.append(node.nullid)
 
     # Generate list of changed files
     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)
 
     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))
             # lfile in manifest if it has not been removed/renamed
             if f in ctx.manifest():
                 if 'l' in ctx.filectx(f).flags():
                     if renamed and renamed[0] in lfiles:
                         raise util.Abort(_('largefile %s becomes symlink') % f)
 
                 # lfile was modified, update standins
                 fullpath = rdst.wjoin(f)
                 lfutil.createdir(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 lfile because lfileness
                 # doesn't change after rename or copy
                 renamed = lfutil.standin(renamed[0])
 
             return context.memfilectx(f, lfiletohash[srcfname], 'l' in
                 fctx.flags(), 'x' in fctx.flags(), renamed)
         else:
             try:
                 fctx = ctx.filectx(f)
             except error.LookupError:
                 raise IOError()
             renamed = fctx.renamed()
             if renamed:
                 renamed = renamed[0]
 
             data = fctx.data()
             if f == '.hgtags':
                 newdata = []
                 for line in data.splitlines():
                     id, name = line.split(' ', 1)
                     newdata.append('%s %s\n' % (node.hex(revmap[node.bin(id)]),
                         name))
                 data = ''.join(newdata)
             return context.memfilectx(f, data, 'l' in fctx.flags(),
                                       'x' in fctx.flags(), renamed)
 
     # Commit
     mctx = context.memctx(rdst, parents, ctx.description(), dstfiles,
                           getfilectx, ctx.user(), ctx.date(), ctx.extra())
     ret = rdst.commitctx(mctx)
     rdst.dirstate.setparents(ret)
     revmap[ctx.node()] = rdst.changelog.tip()
 
 def _islfile(file, ctx, matcher, size):
     '''
     A file is a lfile if it matches a pattern or is over
     the given size.
     '''
     # Never store hgtags or hgignore as lfiles
     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'''
 
     # Don't upload locally. All largefiles are in the system wide cache
     # so the other repo can just get them from there.
     if not files or rdst.local():
         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(_('Missing largefile %s 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:
         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 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):
                 return None # don't try to set the mode or update the dirstate
             ret = 1
         mode = os.stat(absstandin).st_mode
         if mode != os.stat(abslfile).st_mode:
             os.chmod(abslfile, mode)
             ret = 1
     else:
         if os.path.exists(abslfile):
             os.unlink(abslfile)
             ret = -1
     state = repo.dirstate[lfutil.standin(lfile)]
     if state == 'n':
         lfdirstate.normal(lfile)
     elif state == 'r':
         lfdirstate.remove(lfile)
     elif state == 'a':
         lfdirstate.add(lfile)
     elif state == '?':
         lfdirstate.drop(lfile)
     return ret
 
 # -- hg commands declarations ------------------------------------------------
 
 
 cmdtable = {
     'lfconvert': (lfconvert,
                   [('s', 'size', 0, 'All files over this size (in megabytes) '
                   'will be considered largefiles. This can also be specified '
                   'in your hgrc as [largefiles].size.'),
                   ('','tonormal',False,
                       'Convert from a largefiles repo to a normal repo')],
                   _('hg lfconvert SOURCE DEST [FILE ...]')),
     }

hgext/largefiles/lfutil.py

@@ -1,431 +1,445 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.
 
 '''largefiles utility code: must not import other modules in this package.'''
 
 import os
 import errno
 import shutil
 import stat
 import hashlib
 
 from mercurial import dirstate, httpconnection, match as match_, util, scmutil
 from mercurial.i18n import _
 
 shortname = '.hglf'
 longname = 'largefiles'
 
 
 # -- Portability wrappers ----------------------------------------------
 
 def dirstate_walk(dirstate, matcher, unknown=False, ignored=False):
     return dirstate.walk(matcher, [], unknown, ignored)
 
 def repo_add(repo, list):
     add = repo[None].add
     return add(list)
 
 def repo_remove(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 repo_forget(repo, list):
     forget = repo[None].forget
     return forget(list)
 
 def findoutgoing(repo, remote, force):
     from mercurial import discovery
     common, _anyinc, _heads = discovery.findcommonincoming(repo,
         remote, force=force)
     return repo.changelog.findmissing(common)
 
 # -- Private worker functions ------------------------------------------
 
+def getminsize(ui, assumelfiles, opt, default=10):
+    lfsize = opt
+    if not lfsize and assumelfiles:
+        lfsize = ui.config(longname, 'size', default=default)
+    if lfsize:
+        try:
+            lfsize = int(lfsize)
+        except ValueError:
+            raise util.Abort(_('largefiles: size must be an integer, was %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 fall back on copy
         shutil.copyfile(src, dest)
         os.chmod(dest, os.stat(src).st_mode)
 
 def systemcachepath(ui, hash):
     path = ui.config(longname, 'systemcache', None)
     if path:
         path = os.path.join(path, hash)
     else:
         if os.name == 'nt':
             path = os.path.join(os.getenv('LOCALAPPDATA') or \
                 os.getenv('APPDATA'), longname, hash)
         elif os.name == 'posix':
             path = os.path.join(os.getenv('HOME'), '.' + longname, hash)
         else:
             raise util.Abort(_('Unknown operating system: %s\n') % os.name)
     return path
 
 def insystemcache(ui, hash):
     return os.path.exists(systemcachepath(ui, hash))
 
 def findfile(repo, hash):
     if incache(repo, hash):
         repo.ui.note(_('Found %s in cache\n') % hash)
         return cachepath(repo, hash)
     if insystemcache(repo.ui, hash):
         repo.ui.note(_('Found %s in system cache\n') % hash)
         return systemcachepath(repo.ui, hash)
     return None
 
 class largefiles_dirstate(dirstate.dirstate):
     def __getitem__(self, key):
         return super(largefiles_dirstate, self).__getitem__(unixpath(key))
     def normal(self, f):
         return super(largefiles_dirstate, self).normal(unixpath(f))
     def remove(self, f):
         return super(largefiles_dirstate, self).remove(unixpath(f))
     def add(self, f):
         return super(largefiles_dirstate, self).add(unixpath(f))
     def drop(self, f):
         return super(largefiles_dirstate, self).drop(unixpath(f))
     def forget(self, f):
         return super(largefiles_dirstate, self).forget(unixpath(f))
 
 def openlfdirstate(ui, repo):
     '''
     Return a dirstate object that tracks big files: i.e. its root is the
     repo root, but it is saved in .hg/largefiles/dirstate.
     '''
     admin = repo.join(longname)
     opener = scmutil.opener(admin)
     if util.safehasattr(repo.dirstate, '_validate'):
         lfdirstate = largefiles_dirstate(opener, ui, repo.root,
             repo.dirstate._validate)
     else:
         lfdirstate = largefiles_dirstate(opener, ui, repo.root)
 
     # 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.  It also gives us an easy way to forcibly rebuild largefiles
     # state:
     #   rm .hg/largefiles/dirstate && hg status
     # Or even, if things are really messed up:
     #   rm -rf .hg/largefiles && hg status
     if not os.path.exists(os.path.join(admin, 'dirstate')):
         util.makedirs(admin)
         matcher = getstandinmatcher(repo)
         for standin in dirstate_walk(repo.dirstate, matcher):
             lfile = splitstandin(standin)
             hash = readstandin(repo, lfile)
             lfdirstate.normallookup(lfile)
             try:
                 if hash == hashfile(lfile):
                     lfdirstate.normal(lfile)
             except IOError, err:
                 if err.errno != errno.ENOENT:
                     raise
 
         lfdirstate.write()
 
     return lfdirstate
 
 def lfdirstate_status(lfdirstate, repo, rev):
     wlock = repo.wlock()
     try:
         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)
         lfdirstate.write()
     finally:
         wlock.release()
     return (modified, added, removed, missing, unknown, ignored, clean)
 
 def listlfiles(repo, rev=None, matcher=None):
     '''list largefiles in the working copy or 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 incache(repo, hash):
     return os.path.exists(cachepath(repo, hash))
 
 def createdir(dir):
     if not os.path.exists(dir):
         os.makedirs(dir)
 
 def cachepath(repo, hash):
     return repo.join(os.path.join(longname, hash))
 
 def copyfromcache(repo, hash, filename):
     '''copyfromcache copies the specified largefile from the repo or system
     cache to the specified location in the repository.  It will not throw an
     exception on failure, as it 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)))
     shutil.copy(path, repo.wjoin(filename))
     return True
 
 def copytocache(repo, rev, file, uploaded=False):
     hash = readstandin(repo, file)
     if incache(repo, hash):
         return
     copytocacheabsolute(repo, repo.wjoin(file), hash)
 
 def copytocacheabsolute(repo, file, hash):
     createdir(os.path.dirname(cachepath(repo, hash)))
     if insystemcache(repo.ui, hash):
         link(systemcachepath(repo.ui, hash), cachepath(repo, hash))
     else:
         shutil.copyfile(file, cachepath(repo, hash))
         os.chmod(cachepath(repo, hash), os.stat(file).st_mode)
         linktosystemcache(repo, hash)
 
 def linktosystemcache(repo, hash):
     createdir(os.path.dirname(systemcachepath(repo.ui, hash)))
     link(cachepath(repo, hash), systemcachepath(repo.ui, hash))
 
 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 composed_matchfn(f):
         return isstandin(f) and rmatcher.matchfn(splitstandin(f))
     smatcher.matchfn = composed_matchfn
 
     return smatcher
 
 def standin(filename):
     '''Return the repo-relative path to the standin for the specified big
     file.'''
     # Notes:
     # 1) Most callers want an absolute path, but _create_standin() needs
     #    it repo-relative so lfadd() can pass it to repo_add().  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 + '/' + filename.replace(os.sep, '/')
 
 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 = filename.replace(os.sep, '/').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 hhash 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 above.
     infile.close()
 
 def readhash(filename):
     rfile = open(filename, 'rb')
     hash = rfile.read(40)
     rfile.close()
     if len(hash) < 40:
         raise util.Abort(_('bad hash in \'%s\' (only %d bytes long)')
                          % (filename, len(hash)))
     return hash
 
 def writehash(hash, filename, executable):
     util.makedirs(os.path.dirname(filename))
     if os.path.exists(filename):
         os.unlink(filename)
     wfile = open(filename, 'wb')
 
     try:
         wfile.write(hash)
         wfile.write('\n')
     finally:
         wfile.close()
     if os.path.exists(filename):
         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 = hashlib.sha1()
     for chunk in util.filechunkiter(data):
         h.update(chunk)
     return h.hexdigest()
 
 def httpsendfile(ui, filename):
     return httpconnection.httpsendfile(ui, filename, 'rb')
 
 # Convert a path to a unix style path. This is used to give a
 # canonical path to the lfdirstate.
 def unixpath(path):
     return os.path.normpath(path).replace(os.sep, '/')
 
 def islfilesrepo(repo):
     return ('largefiles' in repo.requirements and
             any_(shortname + '/' in f[0] for f in repo.store.datafiles()))
 
 def any_(gen):
     for x in gen:
         if x:
             return True
     return False
 
 class storeprotonotcapable(BaseException):
     def __init__(self, storetypes):
         self.storetypes = storetypes

hgext/largefiles/overrides.py

@@ -1,830 +1,823 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, match as match_, node, \
         archival, error, merge
 from mercurial.i18n import _
 from mercurial.node import hex
 from hgext import rebase
+import lfutil
 
 try:
     from mercurial import scmutil
 except ImportError:
     pass
 
 import lfutil
 import lfcommands
 
 def installnormalfilesmatchfn(manifest):
     '''overrides scmutil.match so that the matcher it returns will ignore all
     largefiles'''
     oldmatch = None # for the closure
     def override_match(repo, pats=[], opts={}, globbed=False,
             default='relpath'):
         match = oldmatch(repo, 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)
         orig_matchfn = m.matchfn
         m.matchfn = lambda f: notlfile(f) and orig_matchfn(f) or None
         return m
     oldmatch = installmatchfn(override_match)
 
 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)
 
 # -- 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 lfiles. Then making a new
 # matcher which matches only the normal files and running the original
 # version of add.
 def override_add(orig, ui, repo, *pats, **opts):
     large = opts.pop('large', None)
-
-    lfsize = opts.pop('lfsize', None)
-    if not lfsize and lfutil.islfilesrepo(repo):
-        lfsize = ui.config(lfutil.longname, 'size', default='10')
-    if lfsize:
-        try:
-            lfsize = int(lfsize)
-        except ValueError:
-            raise util.Abort(_('largefiles: size must be an integer, was %s\n')
-                             % lfsize)
+    lfsize = lfutil.getminsize(
+        ui, lfutil.islfilesrepo(repo), opts.pop('lfsize', None))
 
     lfmatcher = None
     if os.path.exists(repo.wjoin(lfutil.shortname)):
         lfpats = ui.config(lfutil.longname, 'patterns', default=())
         if lfpats:
             lfpats = lfpats.split(' ')
             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:
             if large or (lfsize and os.path.getsize(repo.wjoin(f)) >= \
                     lfsize * 1024 * 1024) 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 inbetween 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.repo_add(repo,
                 standins) if f in m.files()]
     finally:
         wlock.release()
 
     installnormalfilesmatchfn(repo[None].manifest())
     result = orig(ui, repo, *pats, **opts)
     restorematchfn()
 
     return (result == 1 or bad) and 1 or 0
 
 def override_remove(orig, ui, repo, *pats, **opts):
     manifest = repo[None].manifest()
     installnormalfilesmatchfn(manifest)
     orig(ui, repo, *pats, **opts)
     restorematchfn()
 
     after, force = opts.get('after'), opts.get('force')
     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
     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: file %s (use -f to force removal)\n')
                     % (m.rel(f), reason))
 
     if force:
         remove, forget = modified + deleted + clean, added
     elif after:
         remove, forget = deleted, []
         warn(modified + added + clean, _('still exists'))
     else:
         remove, forget = deleted + clean, []
         warn(modified, _('is modified'))
         warn(added, _('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:
                 os.unlink(repo.wjoin(f))
                 currentdir = os.path.split(f)[0]
                 while currentdir and not os.listdir(repo.wjoin(currentdir)):
                     os.rmdir(repo.wjoin(currentdir))
                     currentdir = os.path.split(currentdir)[0]
             lfdirstate.remove(f)
         lfdirstate.write()
 
         forget = [lfutil.standin(f) for f in forget]
         remove = [lfutil.standin(f) for f in remove]
         lfutil.repo_forget(repo, forget)
         lfutil.repo_remove(repo, remove, unlink=True)
     finally:
         wlock.release()
 
 def override_status(orig, ui, repo, *pats, **opts):
     try:
         repo.lfstatus = True
         return orig(ui, repo, *pats, **opts)
     finally:
         repo.lfstatus = False
 
 def override_log(orig, ui, repo, *pats, **opts):
     try:
         repo.lfstatus = True
         orig(ui, repo, *pats, **opts)
     finally:
         repo.lfstatus = False
 
 def override_verify(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 override_update(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 lfiles
     # 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)
 
 # Override filemerge to prompt the user about how they wish to merge lfiles.
 # This will handle identical edits, and copy/rename + edit without prompting
 # the user.
 def override_filemerge(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 lfiles.
 # Then it overrides util.copyfile in that function it checks if the destination
 # lfile already exists. It also keeps a list of copied files so that the lfiles
 # can be copied and the dirstate updated.
 def override_copy(orig, ui, repo, pats, opts, rename=False):
     # doesn't remove lfile 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(os.path.relpath('.', repo.getcwd()),
             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 lfiles and run it again
     nonormalfiles = False
     nolfiles = False
     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:
         # 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 override_match(repo, 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(repo, 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)
             orig_matchfn = m.matchfn
             m.matchfn = lambda f: lfutil.isstandin(f) and \
                 lfile(lfutil.splitstandin(f)) and \
                 orig_matchfn(lfutil.splitstandin(f)) or None
             return m
         oldmatch = installmatchfn(override_match)
         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 override_copyfile(src, dest):
                 if lfutil.shortname in src and lfutil.shortname in dest:
                     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 = override_copyfile
             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 lfutil.shortname in dest:
                 srclfile = src.replace(lfutil.shortname, '')
                 destlfile = dest.replace(lfutil.shortname, '')
                 destlfiledir = os.path.dirname(destlfile) or '.'
                 if not os.path.isdir(destlfiledir):
                     os.makedirs(destlfiledir)
                 if rename:
                     os.rename(srclfile, destlfile)
                     lfdirstate.remove(os.path.relpath(srclfile,
                         repo.root))
                 else:
                     util.copyfile(srclfile, destlfile)
                 lfdirstate.add(os.path.relpath(destlfile,
                     repo.root))
         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 lfiles accidentally.  This means we have to keep track of the lfiles
 # that are being reverted so we only pull down the necessary lfiles.
 #
 # Standins are only updated (to match the hash of lfiles) before commits.
 # Update the standins then run the original revert (changing the matcher to hit
 # standins instead of lfiles). Based on the resulting standins update the
 # lfiles. Then return the standins to their proper state
 def override_revert(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.lfdirstate_status(lfdirstate, repo, repo['.'].rev())
         for lfile in modified:
             lfutil.updatestandin(repo, lfutil.standin(lfile))
 
         try:
             ctx = repo[opts.get('rev')]
             oldmatch = None # for the closure
             def override_match(ctxorrepo, pats=[], opts={}, globbed=False,
                     default='relpath'):
                 if util.safehasattr(ctxorrepo, 'match'):
                     ctx0 = ctxorrepo
                 else:
                     ctx0 = ctxorrepo[None]
                 match = oldmatch(ctxorrepo, pats, opts, globbed, default)
                 m = copy.copy(match)
                 def tostandin(f):
                     if lfutil.standin(f) in ctx0 or 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)
                 orig_matchfn = m.matchfn
                 def matchfn(f):
                     if lfutil.isstandin(f):
                         # We need to keep track of what lfiles are being
                         # matched so we know which ones to update later
                         # (otherwise we revert changes to other lfiles
                         # accidentally).  This is repo specific, so duckpunch
                         # the repo object to keep the list of lfiles for us
                         # later.
                         if orig_matchfn(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 orig_matchfn(f)
                 m.matchfn = matchfn
                 return m
             oldmatch = installmatchfn(override_match)
             scmutil.match
             matches = override_match(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 lfiles 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 hg_update(orig, repo, node):
     result = orig(repo, node)
     # XXX check if it worked first
     lfcommands.updatelfiles(repo.ui, repo)
     return result
 
 def hg_clean(orig, repo, node, show_stats=True):
     result = orig(repo, node, show_stats)
     lfcommands.updatelfiles(repo.ui, repo)
     return result
 
 def hg_merge(orig, repo, node, force=None, remind=True):
     result = orig(repo, node, force, remind)
     lfcommands.updatelfiles(repo.ui, repo)
     return result
 
 # When we rebase a repository with remotely changed lfiles, we need
 # to take some extra care so that the lfiles are correctly updated
 # in the working copy
 def override_pull(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'
         result = orig(ui, repo, source, **opts)
     return result
 
 def override_rebase(orig, ui, repo, **opts):
     repo._isrebasing = True
     try:
         orig(ui, repo, **opts)
     finally:
         repo._isrebasing = False
 
 def override_archive(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 lfile 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())
             f = lfutil.splitstandin(f)
 
             def getdatafn():
                 try:
                     fd = open(path, 'rb')
                     return fd.read()
                 finally:
                     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)
             try:
                 sub.archive(repo.ui, archiver, prefix)
             except TypeError:
                 sub.archive(archiver, prefix)
 
     archiver.done()
 
 # If a lfile 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 lfiles were changed. This is
 # used by bisect and backout.
 def override_bailifchanged(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.bail_if_changed so override it to add the check
 def override_fetch(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 override_forget(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.repo_remove(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 override_outgoing(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 override_summary(orig, ui, repo, *pats, **opts):
     orig(ui, repo, *pats, **opts)
 
     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 override_addremove(orig, ui, repo, *pats, **opts):
     # Check if the parent or child has lfiles if they do don't allow it.  If
     # there is a symlink in the manifest then getting the manifest throws an
     # exception catch it and let addremove deal with it. This happens in
     # Mercurial's test test-addremove-symlink
     try:
         manifesttip = set(repo['tip'].manifest())
     except util.Abort:
         manifesttip = set()
     try:
         manifestworking = set(repo[None].manifest())
     except util.Abort:
         manifestworking = set()
 
     # Manifests are only iterable so turn them into sets then union
     for file in manifesttip.union(manifestworking):
         if file.startswith(lfutil.shortname):
             raise util.Abort(
                 _('addremove cannot be run on a repo with largefiles'))
 
     return orig(ui, repo, *pats, **opts)
 
 # Calling purge with --all will cause the lfiles to be deleted.
 # Override repo.status to prevent this from happening.
 def override_purge(orig, ui, repo, *dirs, **opts):
     oldstatus = repo.status
     def override_status(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 = override_status
     orig(ui, repo, *dirs, **opts)
     repo.status = oldstatus
 
 def override_rollback(orig, ui, repo, **opts):
     result = orig(ui, repo, **opts)
     merge.update(repo, node=None, branchmerge=False, force=True,
         partial=lfutil.isstandin)
     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()
     return result