v6: avoid accidental conversion when annex.largefiles is not configured

v6: When annex.largefiles is not configured for a file, running git add or
git commit, or otherwise using git to stage a file will add it to the annex
if the file was in the annex before, and to git otherwise. This is to avoid
accidental conversion.

Note that git-annex add's behavior has not changed, for reasons explained
in the added comment.

Performance: No added overhead when annex.largefiles is configured.
When not configured, there is an added call to catObjectMetaData,
which involves a round trip through git cat-file --batch.
However, the earlier catKeyFile primes the cache for it.

This commit was supported by the NSF-funded DataLad project.

Annex/CatFile.hs

@@ -13,6 +13,7 @@ module Annex.CatFile (
 	catCommit,
 	catObjectDetails,
 	catFileHandle,
+	catObjectMetaData,
 	catFileStop,
 	catKey,
 	catKeyFile,

Annex/FileMatcher.hs

@@ -10,6 +10,7 @@
 module Annex.FileMatcher (
 	GetFileMatcher,
 	checkFileMatcher,
+	checkFileMatcher',
 	checkMatcher,
 	matchAll,
 	preferredContentParser,
@@ -42,17 +43,25 @@ import qualified Data.Set as S
 type GetFileMatcher = FilePath -> Annex (FileMatcher Annex)
 
 checkFileMatcher :: GetFileMatcher -> FilePath -> Annex Bool
-checkFileMatcher getmatcher file = do
+checkFileMatcher getmatcher file = checkFileMatcher' getmatcher file (return True)
-	matcher <- getmatcher file
-	checkMatcher matcher Nothing (AssociatedFile (Just file)) S.empty True
 
-checkMatcher :: FileMatcher Annex -> Maybe Key -> AssociatedFile -> AssumeNotPresent -> Bool -> Annex Bool
+-- | Allows running an action when no matcher is configured for the file.
-checkMatcher matcher mkey afile notpresent d
+checkFileMatcher' :: GetFileMatcher -> FilePath -> Annex Bool -> Annex Bool
-	| isEmpty matcher = return d
+checkFileMatcher' getmatcher file notconfigured = do
+	matcher <- getmatcher file
+	checkMatcher matcher Nothing afile S.empty notconfigured d
+  where
+	afile = AssociatedFile (Just file)
+	-- checkMatcher will never use this, because afile is provided.
+	d = return True
+
+checkMatcher :: FileMatcher Annex -> Maybe Key -> AssociatedFile -> AssumeNotPresent -> Annex Bool -> Annex Bool -> Annex Bool
+checkMatcher matcher mkey afile notpresent notconfigured d
+	| isEmpty matcher = notconfigured
 	| otherwise = case (mkey, afile) of
 		(_, AssociatedFile (Just file)) -> go =<< fileMatchInfo file
 		(Just key, _) -> go (MatchingKey key)
-		_ -> return d
+		_ -> d
   where
 	go mi = matchMrun matcher $ \a -> a notpresent mi
 

CHANGELOG

@@ -11,6 +11,11 @@ git-annex (6.20180808) UNRELEASED; urgency=medium
     pipe. This also avoids git buffering the whole file content in memory.
   * v6: After updating the worktree for an add/drop, update git's index,
     so git status will not show the files as modified.
+  * v6: When annex.largefiles is not configured for a file, running git
+    add or git commit, or otherwise using git to stage a file
+    will add it to the annex if the file was in the annex before,
+    and to git otherwise. This is to avoid accidental conversion.
+    Note that git-annex add's behavior has not changed.
   * v6: Update associated files database when git has staged changes
     to pointer files.
   * v6: Fix some race conditions.

Command/Smudge.hs

@@ -1,6 +1,6 @@
 {- git-annex command
  -
- - Copyright 2015 Joey Hess <id@joeyh.name>
+ - Copyright 2015-2018 Joey Hess <id@joeyh.name>
  -
  - Licensed under the GNU GPL version 3 or higher.
  -}
@@ -18,6 +18,7 @@ import Logs.Location
 import qualified Database.Keys
 import qualified Git.BuildVersion
 import Git.FilePath
+import Git.Ref
 import Backend
 
 import qualified Data.ByteString.Lazy as B
@@ -77,50 +78,59 @@ clean file = do
 		Just k -> do
 			getMoveRaceRecovery k file
 			liftIO $ B.hPut stdout b
-		Nothing -> ifM (shouldAnnex file)
+		Nothing -> go b =<< catKeyFile file
-			( do
-				-- Before git 2.5, failing to consume all
-				-- stdin here would cause a SIGPIPE and
-				-- crash it.
-				-- Newer git catches the signal and
-				-- stops sending, which is much faster.
-				-- (Also, git seems to forget to free memory
-				-- when sending the file, so the less we
-				-- let it send, the less memory it will
-				-- waste.)
-				if Git.BuildVersion.older "2.5"
-					then B.length b `seq` return ()
-					else liftIO $ hClose stdin
-				-- Look up the backend that was used
-				-- for this file before, so that when
-				-- git re-cleans a file its backend does
-				-- not change.
-				currbackend <- maybe Nothing (maybeLookupBackendVariety . keyVariety)
-					<$> catKeyFile file
-				liftIO . emitPointer
-					=<< go
-					=<< (\ld -> ingest' currbackend ld Nothing norestage)
-					=<< lockDown cfg file
-			, liftIO $ B.hPut stdout b
-			)
 	stop
   where
-	go (Just k, _) = do
+	go b oldkey = ifM (shouldAnnex file oldkey)
+		( do
+			-- Before git 2.5, failing to consume all stdin here
+			-- would cause a SIGPIPE and crash it.
+			-- Newer git catches the signal and stops sending,
+			-- which is much faster. (Also, git seems to forget
+			-- to free memory when sending the file, so the
+			-- less we let it send, the less memory it will waste.)
+			if Git.BuildVersion.older "2.5"
+				then B.length b `seq` return ()
+				else liftIO $ hClose stdin
+			-- Look up the backend that was used for this file
+			-- before, so that when git re-cleans a file its
+			-- backend does not change.
+			let oldbackend = maybe Nothing (maybeLookupBackendVariety . keyVariety) oldkey
+			-- Can't restage associated files because git add
+			-- runs this and has the index locked.
+			let norestage = Restage False
+			liftIO . emitPointer
+				=<< postingest
+				=<< (\ld -> ingest' oldbackend ld Nothing norestage)
+				=<< lockDown cfg file
+		, liftIO $ B.hPut stdout b
+		)
+
+	postingest (Just k, _) = do
 		logStatus k InfoPresent
 		return k
-	go _ = error "could not add file to the annex"
+	postingest _ = error "could not add file to the annex"
+
 	cfg = LockDownConfig
 		{ lockingFile = False
 		, hardlinkFileTmp = False
 		}
-	-- Can't restage associated files because git add runs this and has
-	-- the index locked.
-	norestage = Restage False
 
-shouldAnnex :: FilePath -> Annex Bool
+-- New files are annexed as configured by annex.largefiles, with a default
-shouldAnnex file = do
+-- of annexing them.
+-- 
+-- If annex.largefiles is not configured for a file, and a file with its
+-- name is already in the index, preserve its annexed/not annexed state.
+-- This prevents accidental conversions when annex.largefiles is being
+-- set/unset on the fly rather than being set in gitattributes or .git/config.
+shouldAnnex :: FilePath -> Maybe Key -> Annex Bool
+shouldAnnex file moldkey = do
 	matcher <- largeFilesMatcher
-	checkFileMatcher matcher file
+	checkFileMatcher' matcher file whenempty
+  where
+	whenempty = case moldkey of
+		Just _ -> return True
+		Nothing -> isNothing <$> catObjectMetaData (fileRef file)
 
 emitPointer :: Key -> IO ()
 emitPointer = putStr . formatPointer

Logs/PreferredContent.hs

@@ -58,7 +58,7 @@ checkMap getmap mu notpresent mkey afile d = do
 	m <- getmap
 	case M.lookup u m of
 		Nothing -> return d
-		Just matcher -> checkMatcher matcher mkey afile notpresent d
+		Just matcher -> checkMatcher matcher mkey afile notpresent (return d) (return d)
 
 preferredContentMap :: Annex (FileMatcherMap Annex)
 preferredContentMap = maybe (fst <$> preferredRequiredMapsLoad) return

doc/bugs/Too_difficult_if_not_impossible_to_explicitly_add__47__keep_file_under_git___40__not_annex__41___in_v6_without_employing_.gitattributes.mdwn

@@ -105,3 +105,6 @@ index 982793c..8fdffc0 100644
 yeap
 
 [[!meta author=yoh]]
+
+> [[done]]; clean filter defaults to preserving git/annex state of file.
+> --[[Joey]]

doc/bugs/Too_difficult_if_not_impossible_to_explicitly_add__47__keep_file_under_git___40__not_annex__41___in_v6_without_employing_.gitattributes/comment_12_db64ae67b8aa974bd6d00aad005058ac._comment

@@ -0,0 +1,31 @@
+[[!comment format=mdwn
+ username="joey"
+ subject="""comment 12"""
+ date="2018-08-27T18:08:24Z"
+ content="""
+Hmm, I think if annex.largefiles is configured, it should be honored. So,
+the ways to convert documented on
+<https://git-annex.branchable.com/tips/largefiles/> will work
+with only a minor modification.
+
+If annex.largefiles is not configured, it can check if the file was annexed
+or not before, and maintain the status quo.
+
+At least to start with I am only going to do it in the clean filter, so
+`git annex add` behavior won't change, it default to annexing. It may make
+sense to make `git annex add` consistent with `git add`, but that would
+also need to affect v5 repositories, probably, and I don't want to entangle
+a v6 bug fix with a v5 behavior change. Also, the risk of a
+`git annex add` with the wrong annex.largefiles temp setting
+is much smaller than the risk of forgetting to temp set annex.largefiles
+when running `git commit -a`. Also, I seem to remember another todo item
+discussing making this change to `git annex add` and would need to revisit
+the thinking in that.
+
+Note that a .gitattributes file might only configure largefiles for eg,
+"*.c" and not for other files, thus implicitly accepting the default
+largefiles for the rest. Should then `Makefile` be treated as having
+largefiles configured or not? I lean toward treating it as not configured,
+because then when the user temporarily overrides largefiles to add `Makefile`
+to git, a modification won't accidentially go to the annex.
+"""]]

doc/bugs/committing_files_into_git_doesn__39__t_work_with_explicitly_given_paths_in_V6.mdwn

@@ -45,3 +45,6 @@ operating system: linux x86_64
 
 
 [[ben]]
+
+> [[done]]; clean filter defaults to preserving git/annex state of file.
+> --[[Joey]]

doc/tips/largefiles.mdwn

@@ -128,7 +128,7 @@ When you have a file that is currently stored in the annex, and you want to
 convert that to be stored in git, here's how to accomplish that:
 
 	git annex unlock file
-	git add file
+	git -c annex.largefiles=nothing add file
 	git commit -n file
 
 You can modify the file after unlocking it and before adding it to