include locked files in the keys database associated files

Before only unlocked files were included.

The initial scan now scans for locked as well as unlocked files. This
does mean it gets a little bit slower, although I optimised it as well
as I think it can be.

reconcileStaged changed to diff from the current index to the tree of
the previous index. This lets it handle deletions as well, removing
associated files for both locked and unlocked files, which did not
always happen before.

On upgrade, there will be no recorded previous tree, so it will diff
from the empty tree to current index, and so will fully populate the
associated files, as well as removing any stale associated files
that were present due to them not being removed before.

reconcileStaged now does a bit more work. Most of the time, this will
just be due to running more often, after some change is made to the
index, and since there will be few changes since the last time, it will
not be a noticable overhead. What may turn out to be a noticable
slowdown is after changing to a branch, it has to go through the diff
from the previous index to the new one, and if there are lots of
changes, that could take a long time. Also, after adding a lot of files,
or deleting a lot of files, or moving a large subdirectory, etc.

Command.Lock used removeAssociatedFile, but now that's wrong because a
newly locked file still needs to have its associated file tracked.

Command.Rekey used removeAssociatedFile when the file was unlocked.
It could remove it also when it's locked, but it is not really
necessary, because it changes the index, and so the next time git-annex
run and accesses the keys db, reconcileStaged will run and update it.

There are probably several other places that use addAssociatedFile and
don't need to any more for similar reasons. But there's no harm in
keeping them, and it probably is a good idea to, if only to support
mixing this with older versions of git-annex.

However, mixing this and older versions does risk reconcileStaged not
running, if the older version already ran it on a given index state. So
it's not a good idea to mix versions. This problem could be dealt with
by changing the name of the gitAnnexKeysDbIndexCache, but that would
leave the old file dangling, or it would need to keep trying to remove
it.

Annex/Init.hs

@@ -134,8 +134,8 @@ initialize' mversion = checkInitializeAllowed $ do
 		else deconfigureSmudgeFilter
 	unlessM isBareRepo $ do
 		when supportunlocked $ do
-			showSideAction "scanning for unlocked files"
-			scanUnlockedFiles
+			showSideAction "scanning for annexed files"
+			scanAnnexedFiles
 		hookWrite postCheckoutHook
 		hookWrite postMergeHook
 	AdjustedBranch.checkAdjustedClone >>= \case

Annex/WorkTree.hs

@@ -66,19 +66,19 @@ whenAnnexed a file = ifAnnexed file (a file) (return Nothing)
 ifAnnexed :: RawFilePath -> (Key -> Annex a) -> Annex a -> Annex a
 ifAnnexed file yes no = maybe no yes =<< lookupKey file
 
-{- Find all unlocked files and update the keys database for them. 
+{- Find all annexed files and update the keys database for them.
  - 
  - This is expensive, and so normally the associated files are updated
  - incrementally when changes are noticed. So, this only needs to be done
- - when initializing/upgrading repository.
+ - when initializing/upgrading a repository.
  -
- - Also, the content for the unlocked file may already be present as
+ - Also, the content for an unlocked file may already be present as
  - an annex object. If so, populate the pointer file with it.
  - But if worktree file does not have a pointer file's content, it is left
  - as-is.
  -}
-scanUnlockedFiles :: Annex ()
-scanUnlockedFiles = whenM (inRepo Git.Ref.headExists <&&> not <$> isBareRepo) $ do
+scanAnnexedFiles :: Annex ()
+scanAnnexedFiles = whenM (inRepo Git.Ref.headExists <&&> not <$> isBareRepo) $ do
 	dropold <- liftIO $ newMVar $ 
 		Database.Keys.runWriter $
 			liftIO . Database.Keys.SQL.dropAllAssociatedFiles
@@ -87,9 +87,10 @@ scanUnlockedFiles = whenM (inRepo Git.Ref.headExists <&&> not <$> isBareRepo) $
 		(Git.LsTree.LsTreeLong False)
 		Git.Ref.headRef
 	forM_ l $ \i -> 
-		when (isregfile i) $
-			maybe noop (add dropold i)
-				=<< catKey (Git.LsTree.sha i)
+		maybe noop (add dropold i)
+			=<< catKey'
+				(Git.LsTree.sha i)
+				(fromMaybe 0 (Git.LsTree.size i))				
 	liftIO $ void cleanup
   where
 	isregfile i = case Git.Types.toTreeItemType (Git.LsTree.mode i) of
@@ -101,7 +102,7 @@ scanUnlockedFiles = whenM (inRepo Git.Ref.headExists <&&> not <$> isBareRepo) $
 		let tf = Git.LsTree.file i
 		Database.Keys.runWriter $
 			liftIO . Database.Keys.SQL.addAssociatedFileFast k tf
-		whenM (inAnnex k) $ do
+		whenM (pure (isregfile i) <&&> inAnnex k) $ do
 			f <- fromRepo $ fromTopFilePath tf
 			liftIO (isPointerFile f) >>= \case
 				Just k' | k' == k -> do

Command/Lock.hs

@@ -62,7 +62,7 @@ perform file key = do
 	lockdown =<< calcRepo (gitAnnexLocation key)
 	addLink (CheckGitIgnore False) file key
 		=<< withTSDelta (liftIO . genInodeCache file)
-	next $ cleanup file key
+	next $ return True
   where
 	lockdown obj = do
 		ifM (isUnmodified key obj)
@@ -97,10 +97,5 @@ perform file key = do
 
 	lostcontent = logStatus key InfoMissing
 
-cleanup :: RawFilePath -> Key -> CommandCleanup
-cleanup file key = do
-	Database.Keys.removeAssociatedFile key =<< inRepo (toTopFilePath file)
-	return True
-
 errorModified :: a
 errorModified =  giveup "Locking this file would discard any changes you have made to it. Use 'git annex add' to stage your changes. (Or, use --force to override)"

Command/Migrate.hs

@@ -86,7 +86,7 @@ perform file oldkey oldbackend newbackend = go =<< genkey (fastMigrate oldbacken
 			urls <- getUrls oldkey
 			forM_ urls $ \url ->
 				setUrlPresent newkey url
-			next $ Command.ReKey.cleanup file oldkey newkey
+			next $ Command.ReKey.cleanup file newkey
 		, giveup "failed creating link from old to new key"
 		)
 	genkey Nothing = do

Command/ReKey.hs

@@ -15,8 +15,6 @@ import Annex.Link
 import Annex.Perms
 import Annex.ReplaceFile
 import Logs.Location
-import Git.FilePath
-import qualified Database.Keys
 import Annex.InodeSentinal
 import Utility.InodeCache
 import qualified Utility.RawFilePath as R
@@ -79,7 +77,7 @@ perform file oldkey newkey = do
 		, unlessM (Annex.getState Annex.force) $
 			giveup $ fromRawFilePath file ++ " is not available (use --force to override)"
 		)
-	next $ cleanup file oldkey newkey
+	next $ cleanup file newkey
 
 {- Make a hard link to the old key content (when supported),
  - to avoid wasting disk space. -}
@@ -119,8 +117,8 @@ linkKey file oldkey newkey = ifM (isJust <$> isAnnexLink file)
 					LinkAnnexNoop -> True
 	)
 
-cleanup :: RawFilePath -> Key -> Key -> CommandCleanup
-cleanup file oldkey newkey = do
+cleanup :: RawFilePath -> Key -> CommandCleanup
+cleanup file newkey = do
 	ifM (isJust <$> isAnnexLink file)
 		( do
 			-- Update symlink to use the new key.
@@ -131,8 +129,6 @@ cleanup file oldkey newkey = do
 			liftIO $ whenM (isJust <$> isPointerFile file) $
 				writePointerFile file newkey mode
 			stagePointerFile file mode =<< hashPointerFile newkey
-			Database.Keys.removeAssociatedFile oldkey 
-				=<< inRepo (toTopFilePath file)
 		)
 	whenM (inAnnex newkey) $
 		logStatus newkey InfoPresent

Database/Keys.hs

@@ -1,6 +1,6 @@
 {- Sqlite database of information about Keys
  -
- - Copyright 2015-2019 Joey Hess <id@joeyh.name>
+ - Copyright 2015-2021 Joey Hess <id@joeyh.name>
  -
  - Licensed under the GNU AGPL version 3 or higher.
  -}
@@ -44,6 +44,9 @@ import Git.FilePath
 import Git.Command
 import Git.Types
 import Git.Index
+import Git.Sha
+import Git.Branch (writeTree, update')
+import qualified Git.Ref
 import Config.Smudge
 import qualified Utility.RawFilePath as R
 
@@ -191,20 +194,17 @@ removeInodeCache = runWriterIO . SQL.removeInodeCache
 isInodeKnown :: InodeCache -> SentinalStatus -> Annex Bool
 isInodeKnown i s = or <$> runReaderIO ((:[]) <$$> SQL.isInodeKnown i s)
 
-{- Looks at staged changes to find when unlocked files are copied/moved,
- - and updates associated files in the keys database.
+{- Looks at staged changes to annexed files, and updates the keys database,
+ - so that its information is consistent with the state of the repository.
  -
- - Since staged changes can be dropped later, does not remove any
- - associated files; only adds new associated files.
- -
- - This needs to be run before querying the keys database so that
- - information is consistent with the state of the repository.
+ - This is run with a lock held, so only one process can be running this at
+ - a time.
  -
  - To avoid unncessary work, the index file is statted, and if it's not
  - changed since last time this was run, nothing is done.
  -
- - Note that this is run with a lock held, so only one process can be
- - running this at a time.
+ - A tree is generated from the index, and the diff between that tree
+ - and the last processed tree is examined for changes.
  -
  - This also cleans up after a race between eg a git mv and git-annex
  - get/drop/similar. If git moves the file between this being run and the
@@ -233,17 +233,28 @@ reconcileStaged qh = do
 					)
 		Nothing -> noop
   where
-	go cur indexcache = do
-		(l, cleanup) <- inRepo $ pipeNullSplit' diff
-		changed <- procdiff l False
-		void $ liftIO cleanup
-		-- Flush database changes immediately
-		-- so other processes can see them.
-		when changed $
-			liftIO $ H.flushDbQueue qh
-		liftIO $ writeFile indexcache $ showInodeCache cur
+	lastindexref = Ref "refs/annex/last-index"
 
-	diff =
+	go cur indexcache = do
+		oldtree <- fromMaybe emptyTree
+			<$> inRepo (Git.Ref.sha lastindexref)
+		newtree <- inRepo writeTree
+		when (oldtree /= newtree) $ do
+			(l, cleanup) <- inRepo $ pipeNullSplit' $
+				diff oldtree newtree
+			changed <- procdiff l False
+			void $ liftIO cleanup
+			-- Flush database changes immediately
+			-- so other processes can see them.
+			when changed $
+				liftIO $ H.flushDbQueue qh
+			liftIO $ writeFile indexcache $ showInodeCache cur
+			-- Storing the tree in a ref makes sure it does not
+			-- get garbage collected, and is available to diff
+			-- against next time.
+			inRepo $ update' lastindexref newtree
+	
+	diff oldtree newtree =
 		-- Avoid running smudge or clean filters, since we want the
 		-- raw output, and they would block trying to access the
 		-- locked database. The --raw normally avoids git diff
@@ -253,43 +264,49 @@ reconcileStaged qh = do
 		-- (The -G option may make it be used otherwise.)
 		[ Param "-c", Param "diff.external="
 		, Param "diff"
-		, Param "--cached"
 		, Param "--raw"
 		, Param "-z"
 		, Param "--no-abbrev"
-		-- Optimization: Only find pointer files. This is not
-		-- perfect. A file could start with this and not be a
-		-- pointer file. And a pointer file that is replaced with
-		-- a non-pointer file will match this.
-		, Param $ "-G^" ++ fromRawFilePath (toInternalGitPath $
+		-- Optimization: Limit to pointer files and annex symlinks.
+		-- This is not perfect. A file could contain with this and not
+		-- be a pointer file. And a pointer file that is replaced with
+		-- a non-pointer file will match this. This is only a
+		-- prefilter so that's ok.
+		, Param $ "-G" ++ fromRawFilePath (toInternalGitPath $
 			P.pathSeparator `S.cons` objectDir')
-		-- Don't include files that were deleted, because this only
-		-- wants to update information for files that are present
-		-- in the index.
-		, Param "--diff-filter=AMUT"
 		-- Disable rename detection.
 		, Param "--no-renames"
 		-- Avoid other complications.
 		, Param "--ignore-submodules=all"
 		, Param "--no-ext-diff"
+		, Param (fromRef oldtree)
+		, Param (fromRef newtree)
 		]
 	
 	procdiff (info:file:rest) changed
 		| ":" `S.isPrefixOf` info = case S8.words info of
-			(_colonsrcmode:dstmode:_srcsha:dstsha:_change:[])
-				-- Only want files, not symlinks
-				| dstmode /= fmtTreeItemType TreeSymlink -> do
-					maybe noop (reconcile (asTopFilePath file))
-						=<< catKey (Ref dstsha)
-					procdiff rest True
-				| otherwise -> procdiff rest changed
+			(_colonsrcmode:dstmode:srcsha:dstsha:_change:[]) -> do
+				removed <- catKey (Ref srcsha) >>= \case
+					Just oldkey -> do
+						liftIO $ SQL.removeAssociatedFile oldkey
+							(asTopFilePath file)
+							(SQL.WriteHandle qh)
+						return True
+					Nothing -> return False
+				added <- catKey (Ref dstsha) >>= \case
+					Just key -> do
+						liftIO $ SQL.addAssociatedFile key
+							(asTopFilePath file)
+							(SQL.WriteHandle qh)
+						when (dstmode /= fmtTreeItemType TreeSymlink) $
+							reconcilerace (asTopFilePath file) key
+						return True
+					Nothing -> return False
+				procdiff rest (changed || removed || added)
 			_ -> return changed -- parse failed
 	procdiff _ changed = return changed
 
-	-- Note that database writes done in here will not necessarily 
-	-- be visible to database reads also done in here.
-	reconcile file key = do
-		liftIO $ SQL.addAssociatedFileFast key file (SQL.WriteHandle qh)
+	reconcilerace file key = do
 		caches <- liftIO $ SQL.getInodeCaches key (SQL.ReadHandle qh)
 		keyloc <- calcRepo (gitAnnexLocation key)
 		keypopulated <- sameInodeCache keyloc caches

Upgrade/V5.hs

@@ -47,7 +47,7 @@ upgrade automatic = flip catchNonAsync onexception $ do
 		, do
 			checkGitVersionForIndirectUpgrade
 		)
-	scanUnlockedFiles
+	scanAnnexedFiles
 	configureSmudgeFilter
 	-- Inode sentinal file was only used in direct mode and when
 	-- locking down files as they were added. In v6, it's used more

doc/todo/Avoid_lengthy___34__Scanning_for_unlocked_files_...__34__/comment_4_2cb31617bb7003c5bf0e5def358da0e4._comment

@@ -9,7 +9,12 @@ If most of the files are locked, that would actually make the scan
 somewhere around twice as slow as it currently is. So not a worthwhile
 optimisation.
 
-And I don't see much else there that could be optimised. Possibly the
+Update: Now that the scan also scans for locked files to make the
+associated files include information about them, the catKey optimisation
+did make sense. Unfortunately, that does mean this scan got a little bit
+slower still, since it has to use git ls-tree --long.
+
+I don't see much else there that could be optimised. Possibly the
 ls-tree parser could be made faster but it's already using attoparsec
 so unlikely to be many gains.
 """]]