deal better with clock skew situations, using vector clocks

* Deal with clock skew, both forwards and backwards, when logging
  information to the git-annex branch.
* GIT_ANNEX_VECTOR_CLOCK can now be set to a fixed value (eg 1)
  rather than needing to be advanced each time a new change is made.
* Misuse of GIT_ANNEX_VECTOR_CLOCK will no longer confuse git-annex.

When changing a file in the git-annex branch, the vector clock to use is now
determined by first looking at the current time (or GIT_ANNEX_VECTOR_CLOCK
when set), and comparing it to the newest vector clock already in use in
that file. If a newer time stamp was already in use, advance it forward by
a second instead.

When the clock is set to a time in the past, this avoids logging with
an old timestamp, which would risk that log line later being ignored in favor
of "newer" line that is really not newer.

When a log entry has been made with a clock that was set far ahead in the
future, this avoids newer information being logged with an older timestamp
and so being ignored in favor of that future-timestamped information.
Once all clocks get fixed, this will result in the vector clocks being
incremented, until finally enough time has passed that time gets back ahead
of the vector clock value, and then it will return to usual operation.

(This latter situation is not ideal, but it seems the best that can be done.
The issue with it is, since all writers will be incrementing the last
vector clock they saw, there's no way to tell when one writer made a write
significantly later in time than another, so the earlier write might
arbitrarily be picked when merging. This problem is why git-annex uses
timestamps in the first place, rather than pure vector clocks.)

Advancing forward by 1 second is somewhat arbitrary. setDead
advances a timestamp by just 1 picosecond, and the vector clock could
too. But then it would interfere with setDead, which wants to be
overrulled by any change. So it could use 2 picoseconds or something,
but that seems weird. It could just as well advance it forward by a
minute or whatever, but then it would be harder for real time to catch
up with the vector clock when forward clock slew had happened.

A complication is that many log files contain several different peices of
information, and it may be best to only use vector clocks for the same peice
of information. For example, a key's location log file contains
InfoPresent/InfoMissing for each UUID, and it only looks at the vector
clocks for the UUID that is being changed, and not other UUIDs.

Although exactly where the dividing line is can be hard to determine.
Consider metadata logs, where a field "tag" can have multiple values set
at different times. Should it advance forward past the last tag?
Probably. What about when a different field is set, should it look at
the clocks of other fields? Perhaps not, but currently it does, and
this does not seems like it will cause any problems.

Another one I'm not entirely sure about is the export log, which is
keyed by (fromuuid, touuid). So if multiple repos are exporting to the
same remote, different vector clocks can be used for that remote.
It looks like that's probably ok, because it does not try to determine
what order things occurred when there was an export conflict.

Sponsored-by: Jochen Bartl on Patreon

Annex.hs

@@ -200,7 +200,7 @@ data AnnexState = AnnexState
 	, cachedgitenv :: Maybe (AltIndexFile, FilePath, [(String, String)])
 	, urloptions :: Maybe UrlOptions
 	, insmudgecleanfilter :: Bool
-	, getvectorclock :: IO VectorClock
+	, getvectorclock :: IO CandidateVectorClock
 	}
 
 newAnnexState :: GitConfig -> Git.Repo -> IO AnnexState

Annex/VectorClock.hs

@@ -1,9 +1,11 @@
 {- git-annex vector clocks
  -
- - We don't have a way yet to keep true distributed vector clocks.
+ - These are basically a timestamp. However, when logging a new
- - The next best thing is a timestamp.
+ - value, if the old value has a vector clock that is the same or greater
+ - than the current vector clock, the old vector clock is incremented.
+ - This way, clock skew does not cause confusion.
  -
- - Copyright 2017-2020 Joey Hess <id@joeyh.name>
+ - Copyright 2017-2021 Joey Hess <id@joeyh.name>
  -
  - Licensed under the GNU AGPL version 3 or higher.
  -}
@@ -21,10 +23,12 @@ import Utility.TimeStamp
 import Data.ByteString.Builder
 import qualified Data.Attoparsec.ByteString.Lazy as A
 
-currentVectorClock :: Annex VectorClock
+currentVectorClock :: Annex CandidateVectorClock
 currentVectorClock = liftIO =<< Annex.getState Annex.getvectorclock
 
--- Runs the action and uses the same vector clock throughout.
+-- Runs the action and uses the same vector clock throughout,
+-- except when it's necessary to use a newer one due to a past value having
+-- a newer vector clock.
 --
 -- When the action modifies several files in the git-annex branch,
 -- this can cause less space to be used, since the same vector clock
@@ -52,6 +56,19 @@ reuseVectorClockWhile = bracket setup cleanup . const
 	use vc = Annex.changeState $ \s ->
 		s { Annex.getvectorclock = vc }
 
+-- Convert a candidate vector clock in to the final one to use,
+-- advancing it if necessary when necessary to get ahead of a previously
+-- used vector clock.
+advanceVectorClock :: CandidateVectorClock -> [VectorClock] -> VectorClock
+advanceVectorClock (CandidateVectorClock c) [] = VectorClock c
+advanceVectorClock (CandidateVectorClock c) prevs
+	| prev >= VectorClock c = case prev of
+		VectorClock v -> VectorClock (v + 1)
+		Unknown -> VectorClock c
+	| otherwise = VectorClock c
+  where
+	prev = maximum prevs
+
 formatVectorClock :: VectorClock -> String
 formatVectorClock Unknown = "0"
 formatVectorClock (VectorClock t) = show t

Annex/VectorClock/Utility.hs

@@ -13,11 +13,11 @@ import Types.VectorClock
 import Utility.Env
 import Utility.TimeStamp
 
-startVectorClock :: IO (IO VectorClock)
+startVectorClock :: IO (IO CandidateVectorClock)
 startVectorClock = go =<< getEnv "GIT_ANNEX_VECTOR_CLOCK"
   where
 	go Nothing = timebased
 	go (Just s) = case parsePOSIXTime s of
-		Just t -> return (pure (VectorClock t))
+		Just t -> return (pure (CandidateVectorClock t))
 		Nothing -> timebased
-	timebased = return (VectorClock <$> getPOSIXTime)
+	timebased = return (CandidateVectorClock <$> getPOSIXTime)

CHANGELOG

@@ -1,5 +1,10 @@
 git-annex (8.20210804) UNRELEASED; urgency=medium
 
+  * Deal with clock skew, both forwards and backwards, when logging
+    information to the git-annex branch.
+  * GIT_ANNEX_VECTOR_CLOCK can now be set to a fixed value (eg 1)
+    rather than needing to be advanced each time a new change is made.
+  * Misuse of GIT_ANNEX_VECTOR_CLOCK will no longer confuse git-annex.
   * add: When adding a dotfile, avoid treating its name as an extension.
 
  -- Joey Hess <id@joeyh.name>  Tue, 03 Aug 2021 12:22:45 -0400

Command/MetaData.hs

@@ -98,12 +98,12 @@ seek o = case batchOption o of
 			)
 		_ -> giveup "--batch is currently only supported in --json mode"
 
-start :: VectorClock -> MetaDataOptions -> SeekInput -> RawFilePath -> Key -> CommandStart
+start :: CandidateVectorClock -> MetaDataOptions -> SeekInput -> RawFilePath -> Key -> CommandStart
 start c o si file k = startKeys c o (si, k, mkActionItem (k, afile))
   where
 	afile = AssociatedFile (Just file)
 
-startKeys :: VectorClock -> MetaDataOptions -> (SeekInput, Key, ActionItem) -> CommandStart
+startKeys :: CandidateVectorClock -> MetaDataOptions -> (SeekInput, Key, ActionItem) -> CommandStart
 startKeys c o (si, k, ai) = case getSet o of
 	Get f -> startingCustomOutput k $ do
 		l <- S.toList . currentMetaDataValues f <$> getCurrentMetaData k
@@ -113,7 +113,7 @@ startKeys c o (si, k, ai) = case getSet o of
 	_ -> starting "metadata" ai si $
 		perform c o k
 
-perform :: VectorClock -> MetaDataOptions -> Key -> CommandPerform
+perform :: CandidateVectorClock -> MetaDataOptions -> Key -> CommandPerform
 perform c o k = case getSet o of
 	Set ms -> do
 		oldm <- getCurrentMetaData k

Logs/Export.hs

@@ -97,9 +97,8 @@ recordExportUnderway remoteuuid ec = do
 	hereuuid <- getUUID
 	let ep = ExportParticipants { exportFrom = hereuuid, exportTo = remoteuuid }
 	let exported = mkExported (newTreeish ec) []
-	Annex.Branch.change
+	let ru = Annex.Branch.RegardingUUID [remoteuuid, hereuuid]
-		(Annex.Branch.RegardingUUID [remoteuuid, hereuuid]) 
+	Annex.Branch.change ru exportLog $ 
-		exportLog $
 		buildExportLog
 			. changeMapLog c ep exported 
 			. M.mapWithKey (updateForExportChange remoteuuid ec c hereuuid)

Logs/Export/Pure.hs

@@ -23,6 +23,7 @@ module Logs.Export.Pure (
 ) where
 
 import Annex.Common
+import Annex.VectorClock
 import qualified Git
 import Logs.MapLog
 
@@ -110,7 +111,9 @@ exportedParser = Exported <$> refparser <*> many refparser
 -- This way, when multiple repositories are exporting to
 -- the same special remote, there's no conflict as long as they move
 -- forward in lock-step.
-updateForExportChange :: UUID -> ExportChange -> VectorClock -> UUID -> ExportParticipants -> LogEntry Exported -> LogEntry Exported
+updateForExportChange :: UUID -> ExportChange -> CandidateVectorClock -> UUID -> ExportParticipants -> LogEntry Exported -> LogEntry Exported
-updateForExportChange remoteuuid ec c hereuuid ep le@(LogEntry _ exported@(Exported { exportedTreeish = t }))
+updateForExportChange remoteuuid ec c hereuuid ep le@(LogEntry lc exported@(Exported { exportedTreeish = t }))
 	| hereuuid == exportFrom ep || remoteuuid /= exportTo ep || t `notElem` oldTreeish ec = le
-	| otherwise = LogEntry c (exported { exportedTreeish = newTreeish ec })
+	| otherwise = LogEntry c' (exported { exportedTreeish = newTreeish ec })
+  where
+	c' = advanceVectorClock c [lc]

Logs/Location.hs

@@ -8,7 +8,7 @@
  - Repositories record their UUID and the date when they --get or --drop
  - a value.
  - 
- - Copyright 2010-2018 Joey Hess <id@joeyh.name>
+ - Copyright 2010-2021 Joey Hess <id@joeyh.name>
  -
  - Licensed under the GNU AGPL version 3 or higher.
  -}
@@ -61,14 +61,14 @@ logStatusAfter key a = ifM a
 
 {- Log a change in the presence of a key's value in a repository. -}
 logChange :: Key -> UUID -> LogStatus -> Annex ()
-logChange = logChange' logNow
+logChange key u@(UUID _) s = do
-
-logChange' :: (LogStatus -> LogInfo -> Annex LogLine) -> Key -> UUID -> LogStatus -> Annex ()
-logChange' mklog key u@(UUID _) s = do
 	config <- Annex.getGitConfig
-	maybeAddLog (Annex.Branch.RegardingUUID [u]) (locationLogFile config key)
+	maybeAddLog
-		=<< mklog s (LogInfo (fromUUID u))
+		(Annex.Branch.RegardingUUID [u])
-logChange' _ _ NoUUID _ = noop
+		(locationLogFile config key)
+		s
+		(LogInfo (fromUUID u))
+logChange _ NoUUID _ = noop
 
 {- Returns a list of repository UUIDs that, according to the log, have
  - the value of a key. -}
@@ -107,6 +107,9 @@ checkDead key = do
  - 
  - Changes all logged lines for the key, in any location, that are
  - currently InfoMissing, to be InfoDead.
+ - 
+ - The vector clock in the log is updated minimally, so that any
+ - other location log changes are guaranteed to overrule this.
  -}
 setDead :: Key -> Annex ()
 setDead key = do
@@ -117,18 +120,12 @@ setDead key = do
   where
 	go logfile l = 
 		let u = toUUID (fromLogInfo (info l))
-		in addLog (Annex.Branch.RegardingUUID [u]) logfile (setDead' l)
+		    c = case date l of
-
+			VectorClock v -> CandidateVectorClock $
-{- Note that the timestamp in the log is updated minimally, so that this
+				v + realToFrac (picosecondsToDiffTime 1)
- - can be overruled by other location log changes. -}
+			Unknown -> CandidateVectorClock 0
-setDead' :: LogLine -> LogLine
+		in addLog' (Annex.Branch.RegardingUUID [u]) logfile InfoDead
-setDead' l = l
+			(info l) c
-	{ status = InfoDead
-	, date = case date l of
-		VectorClock c -> VectorClock $
-			c + realToFrac (picosecondsToDiffTime 1)
-		Unknown -> Unknown
-	}
 
 data Unchecked a = Unchecked (Annex (Maybe a))
 

Logs/MapLog.hs

@@ -67,8 +67,12 @@ mapLogParser fieldparser valueparser = M.fromListWith best <$> parseLogLines go
 		A.endOfInput
 		return (f, LogEntry c v)
 
-changeMapLog :: Ord f => VectorClock -> f -> v -> MapLog f v -> MapLog f v
+changeMapLog :: Ord f => CandidateVectorClock -> f -> v -> MapLog f v -> MapLog f v
-changeMapLog c f v = M.insert f $ LogEntry c v
+changeMapLog c f v m = M.insert f (LogEntry c' v) m
+  where
+	c' = case M.lookup f m of
+		Nothing -> advanceVectorClock c []
+		Just old -> advanceVectorClock c [changed old]
 
 {- Only add an LogEntry if it's newer (or at least as new as) than any
  - existing LogEntry for a field. -}

Logs/MetaData.hs

@@ -105,22 +105,23 @@ addMetaData' :: Annex.Branch.RegardingUUID -> (GitConfig -> Key -> RawFilePath)
 addMetaData' ru getlogfile k metadata = 
 	addMetaDataClocked' ru getlogfile k metadata =<< currentVectorClock
 
-{- Reusing the same VectorClock when making changes to the metadata
+{- Reusing the same CandidateVectorClock when making changes to the metadata
  - of multiple keys is a nice optimisation. The same metadata lines
  - will tend to be generated across the different log files, and so
  - git will be able to pack the data more efficiently. -}
-addMetaDataClocked :: Key -> MetaData -> VectorClock -> Annex ()
+addMetaDataClocked :: Key -> MetaData -> CandidateVectorClock -> Annex ()
 addMetaDataClocked = addMetaDataClocked' (Annex.Branch.RegardingUUID []) metaDataLogFile
 
-addMetaDataClocked' :: Annex.Branch.RegardingUUID -> (GitConfig -> Key -> RawFilePath) -> Key -> MetaData -> VectorClock -> Annex ()
+addMetaDataClocked' :: Annex.Branch.RegardingUUID -> (GitConfig -> Key -> RawFilePath) -> Key -> MetaData -> CandidateVectorClock -> Annex ()
 addMetaDataClocked' ru getlogfile k d@(MetaData m) c
 	| d == emptyMetaData = noop
 	| otherwise = do
 		config <- Annex.getGitConfig
-		Annex.Branch.change ru (getlogfile config k) $
+		Annex.Branch.change ru (getlogfile config k) $ \b ->
-			buildLog . simplifyLog 
+			let s = parseLog b
-				. S.insert (LogEntry c metadata)
+			    c' = advanceVectorClock c (map changed (S.toList s))
-				. parseLog
+			    ent = LogEntry c' metadata
+			in buildLog $ simplifyLog $ S.insert ent s
   where
 	metadata = MetaData $ M.filterWithKey (\f _ -> not (isLastChangedField f)) m
 

Logs/Presence.hs

@@ -6,7 +6,7 @@
  - A line of the log will look like: "date N INFO"
  - Where N=1 when the INFO is present, 0 otherwise.
  - 
- - Copyright 2010-2014 Joey Hess <id@joeyh.name>
+ - Copyright 2010-2021 Joey Hess <id@joeyh.name>
  -
  - Licensed under the GNU AGPL version 3 or higher.
  -}
@@ -14,9 +14,9 @@
 module Logs.Presence (
 	module X,
 	addLog,
+	addLog',
 	maybeAddLog,
 	readLog,
-	logNow,
 	currentLog,
 	currentLogInfo,
 	historicalLogInfo,
@@ -28,32 +28,43 @@ import Annex.VectorClock
 import qualified Annex.Branch
 import Git.Types (RefDate)
 
-{- Adds a LogLine to the log, removing any LogLines that are obsoleted by
+{- Adds to the log, removing any LogLines that are obsoleted. -}
- - adding it. -}
+addLog :: Annex.Branch.RegardingUUID -> RawFilePath -> LogStatus -> LogInfo -> Annex ()
-addLog :: Annex.Branch.RegardingUUID -> RawFilePath -> LogLine -> Annex ()
+addLog ru file logstatus loginfo = 
-addLog ru file line = Annex.Branch.change ru file $ \b ->
+	addLog' ru file logstatus loginfo =<< currentVectorClock
-	buildLog $ compactLog (line : parseLog b)
+
+addLog' :: Annex.Branch.RegardingUUID -> RawFilePath -> LogStatus -> LogInfo -> CandidateVectorClock -> Annex ()
+addLog' ru file logstatus loginfo c = 
+	Annex.Branch.change ru file $ \b ->
+		let old = parseLog b
+		    line = genLine logstatus loginfo c old
+		in buildLog $ compactLog (line : old)
 
 {- When a LogLine already exists with the same status and info, but an
  - older timestamp, that LogLine is preserved, rather than updating the log
  - with a newer timestamp.
  -}
-maybeAddLog :: Annex.Branch.RegardingUUID -> RawFilePath -> LogLine -> Annex ()
+maybeAddLog :: Annex.Branch.RegardingUUID -> RawFilePath -> LogStatus -> LogInfo -> Annex ()
-maybeAddLog ru file line = Annex.Branch.maybeChange ru file $ \s -> do
+maybeAddLog ru file logstatus loginfo = do
-	m <- insertNewStatus line $ logMap $ parseLog s
+	c <- currentVectorClock
-	return $ buildLog $ mapLog m
+	Annex.Branch.maybeChange ru file $ \b ->
+		let old = parseLog b
+		    line = genLine logstatus loginfo c old
+		in do
+			m <- insertNewStatus line $ logMap old
+			return $ buildLog $ mapLog m
+
+genLine :: LogStatus -> LogInfo -> CandidateVectorClock -> [LogLine] -> LogLine
+genLine logstatus loginfo c old = LogLine c' logstatus loginfo
+  where
+	oldcs = map date (filter (\l -> info l == loginfo) old)
+	c' = advanceVectorClock c oldcs
 
 {- Reads a log file.
  - Note that the LogLines returned may be in any order. -}
 readLog :: RawFilePath -> Annex [LogLine]
 readLog = parseLog <$$> Annex.Branch.get
 
-{- Generates a new LogLine with the current time. -}
-logNow :: LogStatus -> LogInfo -> Annex LogLine
-logNow s i = do
-	c <- currentVectorClock
-	return $ LogLine c s i
-
 {- Reads a log and returns only the info that is still in effect. -}
 currentLogInfo :: RawFilePath -> Annex [LogInfo]
 currentLogInfo file = map info <$> currentLog file

Logs/Presence/Pure.hs

@@ -123,4 +123,5 @@ instance Arbitrary LogLine where
 			(\b -> C8.notElem '\n' b && C8.notElem '\r' b)
 
 prop_parse_build_presence_log :: [LogLine] -> Bool
-prop_parse_build_presence_log l = parseLog (toLazyByteString (buildLog l)) == l
+prop_parse_build_presence_log l =
+	parseLog (toLazyByteString (buildLog l)) == l

Logs/SingleValue.hs

@@ -1,3 +1,5 @@
+{-# LANGUAGE ScopedTypeVariables #-}
+
 {- git-annex single-value log
  -
  - This is used to store a value in a way that can be union merged.
@@ -6,7 +8,7 @@
  -
  - The line with the newest timestamp wins.
  -
- - Copyright 2014 Joey Hess <id@joeyh.name>
+ - Copyright 2014-2021 Joey Hess <id@joeyh.name>
  -
  - Licensed under the GNU AGPL version 3 or higher.
  -}
@@ -31,8 +33,10 @@ readLog = parseLog <$$> Annex.Branch.get
 getLog :: (Ord v, SingleValueSerializable v) => RawFilePath -> Annex (Maybe v)
 getLog = newestValue <$$> readLog
 
-setLog :: (SingleValueSerializable v) => Annex.Branch.RegardingUUID -> RawFilePath -> v -> Annex ()
+setLog :: (Ord v, SingleValueSerializable v) => Annex.Branch.RegardingUUID -> RawFilePath -> v -> Annex ()
 setLog ru f v = do
 	c <- currentVectorClock
-	let ent = LogEntry c v
+	Annex.Branch.change ru f $ \old ->
-	Annex.Branch.change ru f $ \_old -> buildLog (S.singleton ent)
+		let oldcs = map changed ((parseLog' old) `asTypeOf` [ent])
+		    ent = LogEntry (advanceVectorClock c oldcs) v
+		in buildLog (S.singleton ent)

Logs/SingleValue/Pure.hs

@@ -1,6 +1,6 @@
 {- git-annex single-value log, pure operations
  -
- - Copyright 2014-2019 Joey Hess <id@joeyh.name>
+ - Copyright 2014-2021 Joey Hess <id@joeyh.name>
  -
  - Licensed under the GNU AGPL version 3 or higher.
  -}
@@ -33,13 +33,18 @@ buildLog :: (SingleValueSerializable v) => Log v -> Builder
 buildLog = mconcat . map genline . S.toList
   where
 	genline (LogEntry c v) =
-		buildVectorClock c <> sp <> byteString (serialize v) <> nl
+		buildVectorClock c
+			<> sp 
+			<> byteString (serialize v)
+			<> nl
 	sp = charUtf8 ' '
 	nl = charUtf8 '\n'
 
 parseLog :: (Ord v, SingleValueSerializable v) => L.ByteString -> Log v
-parseLog = S.fromList . fromMaybe [] 
+parseLog = S.fromList . parseLog'
-	. A.maybeResult . A.parse (logParser <* A.endOfInput)
+
+parseLog' :: SingleValueSerializable v => L.ByteString -> [LogEntry v]
+parseLog' = fromMaybe [] . A.maybeResult . A.parse (logParser <* A.endOfInput)
 
 logParser :: SingleValueSerializable v => A.Parser [LogEntry v]
 logParser = parseLogLines $ LogEntry

Logs/Transitions.hs

@@ -51,8 +51,12 @@ describeTransition ForgetDeadRemotes = "forget dead remotes"
 noTransitions :: Transitions
 noTransitions = S.empty
 
-addTransition :: VectorClock -> Transition -> Transitions -> Transitions
+addTransition :: CandidateVectorClock -> Transition -> Transitions -> Transitions
-addTransition c t = S.insert $ TransitionLine c (Right t)
+addTransition c t s = S.insert (TransitionLine c' (Right t)) s
+  where
+	oldcs = map transitionStarted $
+		filter (\l -> transition l == Right t) (S.toList s)
+	c' = advanceVectorClock c oldcs
 
 buildTransitions :: Transitions -> Builder
 buildTransitions = mconcat . map genline . S.elems
@@ -73,9 +77,6 @@ parseTransitionsStrictly source b =
 		then ts
 		else giveup $ "unknown transitions listed in " ++ source ++ "; upgrade git-annex!"
 
-showTransitionLine :: TransitionLine -> String
-showTransitionLine (TransitionLine c t) = unwords [show t, formatVectorClock c]
-
 transitionLineParser :: A.Parser TransitionLine
 transitionLineParser = do
 	t <- (parsetransition <$> A.takeByteString)

Logs/UUIDBased.hs

@@ -53,8 +53,10 @@ buildLogOld :: (v -> Builder) -> Log v -> Builder
 buildLogOld builder = mconcat . map genline . M.toList
   where
 	genline (u, LogEntry c@(VectorClock {}) v) =
-		buildUUID u <> sp <> builder v <> sp <>
+		buildUUID u <> sp <> builder v <> sp
-			byteString "timestamp=" <> buildVectorClock c <> nl
+			<> byteString "timestamp="
+			<> buildVectorClock c
+			<> nl
 	genline (u, LogEntry Unknown v) =
 		buildUUID u <> sp <> builder v <> nl
 	sp = charUtf8 ' '
@@ -89,7 +91,7 @@ buildLogNew = buildMapLog buildUUID
 parseLogNew :: A.Parser v -> L.ByteString -> Log v
 parseLogNew = parseMapLog (toUUID <$> A.takeByteString)
 
-changeLog :: VectorClock -> UUID -> v -> Log v -> Log v
+changeLog :: CandidateVectorClock -> UUID -> v -> Log v -> Log v
 changeLog = changeMapLog
 
 addLog :: UUID -> LogEntry v -> Log v -> Log v

Logs/Web.hs

@@ -61,7 +61,7 @@ setUrlPresent key url = do
 	unless (url `elem` us) $ do
 		config <- Annex.getGitConfig
 		addLog (Annex.Branch.RegardingUUID []) (urlLogFile config key)
-			=<< logNow InfoPresent (LogInfo (encodeBS url))
+			InfoPresent (LogInfo (encodeBS url))
 	-- If the url does not have an OtherDownloader, it must be present
 	-- in the web.
 	case snd (getDownloader url) of
@@ -75,7 +75,7 @@ setUrlMissing key url = do
 	when (url `elem` us) $ do
 		config <- Annex.getGitConfig
 		addLog (Annex.Branch.RegardingUUID []) (urlLogFile config key)
-			=<< logNow InfoMissing (LogInfo (encodeBS url))
+			InfoMissing (LogInfo (encodeBS url))
 		-- If the url was a web url and none of the remaining urls
 		-- for the key are web urls, the key must not be present
 		-- in the web.

Types/VectorClock.hs

@@ -1,9 +1,6 @@
 {- git-annex vector clocks
  -
- - We don't have a way yet to keep true distributed vector clocks.
+ - Copyright 2017-2021 Joey Hess <id@joeyh.name>
- - The next best thing is a timestamp.
- -
- - Copyright 2017-2020 Joey Hess <id@joeyh.name>
  -
  - Licensed under the GNU AGPL version 3 or higher.
  -}
@@ -21,6 +18,11 @@ import Utility.QuickCheck
 data VectorClock = Unknown | VectorClock POSIXTime
 	deriving (Eq, Ord, Show)
 
+-- | This is a candidate value to use in a VectorClock. It
+-- may not be suitable to use this, when a previously used VectorClock
+-- is the same or higher.
+data CandidateVectorClock = CandidateVectorClock POSIXTime
+
 -- Unknown is oldest.
 prop_VectorClock_sane :: Bool
 prop_VectorClock_sane = Unknown < VectorClock 1

doc/design/new_repo_versions.mdwn

@@ -103,10 +103,6 @@ Possible reasons to make changes:
   performance. The disk usage change of this method has not yet been
   quantified.
 
-* Another reason to do it would be improving git-annex to use vector clocks,
-  instead of its current assumption that client's clocks are close enough to
-  accurate. This would presumably change the contents of the files.
-
 * While not a sufficient reason on its own, the best practices for file
   formats in the git-annex branch has evolved over time, and there are some
   files that have unusual formats for historical reasons. Other files have

doc/git-annex.mdwn

@@ -1908,14 +1908,13 @@ These environment variables are used by git-annex when set:
 
   Normally git-annex timestamps lines in the log files committed to the
   git-annex branch. Setting this environment variable to a number
-  will make git-annex use that rather than the current number of seconds
+  will make git-annex use that (or a larger number) 
-  since the UNIX epoch. Note that decimal seconds are supported.
+  rather than the current number of seconds since the UNIX epoch.
+  Note that decimal seconds are supported.
   
   This is only provided for advanced users who either have a better way to
   tell which commit is current than the local clock, or who need to avoid
-  embedding timestamps for policy reasons. Misuse of this environment
+  embedding timestamps for policy reasons.
-  variable can confuse git-annex's book-keeping, sometimes in ways that
-  `git annex fsck` is unable to repair.
 
 * Some special remotes use additional environment variables
   for authentication etc. For example, `AWS_ACCESS_KEY_ID`