92b7b1964d
Warn when adding a annex symlink or pointer file that uses a key that is not known to the repository, to prevent confusion if the user has copied it from some other repository. This commit was sponsored by Jake Vosloo on Patreon.
262 lines
8.5 KiB
Haskell
262 lines
8.5 KiB
Haskell
{- git-annex command
|
|
-
|
|
- Copyright 2015-2019 Joey Hess <id@joeyh.name>
|
|
-
|
|
- Licensed under the GNU AGPL version 3 or higher.
|
|
-}
|
|
|
|
module Command.Smudge where
|
|
|
|
import Command
|
|
import Annex.Content
|
|
import Annex.Link
|
|
import Annex.FileMatcher
|
|
import Annex.Ingest
|
|
import Annex.CatFile
|
|
import Logs.Smudge
|
|
import Logs.Location
|
|
import qualified Database.Keys
|
|
import qualified Git.BuildVersion
|
|
import Git.FilePath
|
|
import Git.Types
|
|
import Git.HashObject
|
|
import qualified Git
|
|
import qualified Git.Ref
|
|
import qualified Annex
|
|
import Backend
|
|
import Utility.Metered
|
|
import Annex.InodeSentinal
|
|
import Utility.InodeCache
|
|
import Config.GitConfig
|
|
|
|
import qualified Data.ByteString as S
|
|
import qualified Data.ByteString.Lazy as L
|
|
|
|
cmd :: Command
|
|
cmd = noCommit $ noMessages $
|
|
command "smudge" SectionPlumbing
|
|
"git smudge filter"
|
|
paramFile (seek <$$> optParser)
|
|
|
|
data SmudgeOptions = UpdateOption | SmudgeOptions
|
|
{ smudgeFile :: FilePath
|
|
, cleanOption :: Bool
|
|
}
|
|
|
|
optParser :: CmdParamsDesc -> Parser SmudgeOptions
|
|
optParser desc = smudgeoptions <|> updateoption
|
|
where
|
|
smudgeoptions = SmudgeOptions
|
|
<$> argument str ( metavar desc )
|
|
<*> switch ( long "clean" <> help "clean filter" )
|
|
updateoption = flag' UpdateOption
|
|
( long "update" <> help "populate annexed worktree files" )
|
|
|
|
seek :: SmudgeOptions -> CommandSeek
|
|
seek (SmudgeOptions f False) = commandAction (smudge f)
|
|
seek (SmudgeOptions f True) = commandAction (clean (toRawFilePath f))
|
|
seek UpdateOption = commandAction update
|
|
|
|
-- Smudge filter is fed git file content, and if it's a pointer to an
|
|
-- available annex object, git expects it to output its content.
|
|
--
|
|
-- However, this does not do that. It outputs the pointer, and records
|
|
-- the filename in the smudge log. Git hooks run after commands like checkout
|
|
-- then run git annex smudge --update which populates the work tree files
|
|
-- with annex content. This is done for several reasons:
|
|
--
|
|
-- * To support annex.thin
|
|
-- * Because git currently buffers the whole object received from the
|
|
-- smudge filter in memory, which is a problem with large files.
|
|
smudge :: FilePath -> CommandStart
|
|
smudge file = do
|
|
b <- liftIO $ L.hGetContents stdin
|
|
case parseLinkTargetOrPointerLazy b of
|
|
Nothing -> noop
|
|
Just k -> do
|
|
topfile <- inRepo (toTopFilePath (toRawFilePath file))
|
|
Database.Keys.addAssociatedFile k topfile
|
|
void $ smudgeLog k topfile
|
|
liftIO $ L.putStr b
|
|
stop
|
|
|
|
-- Clean filter is fed file content on stdin, decides if a file
|
|
-- should be stored in the annex, and outputs a pointer to its
|
|
-- injested content if so. Otherwise, the original content.
|
|
clean :: RawFilePath -> CommandStart
|
|
clean file = do
|
|
b <- liftIO $ L.hGetContents stdin
|
|
ifM fileoutsiderepo
|
|
( liftIO $ L.hPut stdout b
|
|
, do
|
|
-- Avoid a potential deadlock.
|
|
Annex.changeState $ \s -> s
|
|
{ Annex.insmudgecleanfilter = True }
|
|
go b
|
|
)
|
|
stop
|
|
where
|
|
go b = case parseLinkTargetOrPointerLazy b of
|
|
Just k -> do
|
|
addingExistingLink file k $ do
|
|
getMoveRaceRecovery k file
|
|
liftIO $ L.hPut stdout b
|
|
Nothing -> do
|
|
let fileref = Git.Ref.fileRef file
|
|
indexmeta <- catObjectMetaData fileref
|
|
go' b indexmeta =<< catKey' fileref indexmeta
|
|
go' b indexmeta oldkey = ifM (shouldAnnex file indexmeta 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 L.length b `seq` return ()
|
|
else liftIO $ hClose stdin
|
|
|
|
-- Optimization for the case when the file is already
|
|
-- annexed and is unmodified.
|
|
case oldkey of
|
|
Nothing -> doingest oldkey
|
|
Just ko -> ifM (isUnmodifiedCheap ko file)
|
|
( liftIO $ emitPointer ko
|
|
, doingest oldkey
|
|
)
|
|
, liftIO $ L.hPut stdout b
|
|
)
|
|
|
|
doingest oldkey = do
|
|
-- Look up the backend that was used for this file
|
|
-- before, so that when git re-cleans a file its
|
|
-- backend does not change.
|
|
oldbackend <- maybe
|
|
(pure Nothing)
|
|
(maybeLookupBackendVariety . fromKey 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 nullMeterUpdate ld Nothing norestage)
|
|
=<< lockDown cfg (fromRawFilePath file)
|
|
|
|
postingest (Just k, _) = do
|
|
logStatus k InfoPresent
|
|
return k
|
|
postingest _ = error "could not add file to the annex"
|
|
|
|
cfg = LockDownConfig
|
|
{ lockingFile = False
|
|
, hardlinkFileTmpDir = Nothing
|
|
}
|
|
|
|
-- git diff can run the clean filter on files outside the
|
|
-- repository; can't annex those
|
|
fileoutsiderepo = do
|
|
repopath <- liftIO . absPath =<< fromRepo Git.repoPath
|
|
filepath <- liftIO $ absPath file
|
|
return $ not $ dirContains repopath filepath
|
|
|
|
-- If annex.largefiles is configured, matching files are added to the
|
|
-- annex. But annex.gitaddtoannex can be set to false to disable that.
|
|
--
|
|
-- When annex.largefiles is not configured, files are normally not
|
|
-- added to the annex, so will be added to git. But some heuristics
|
|
-- are used to avoid bad behavior:
|
|
--
|
|
-- If the file is annexed in the index, keep it annexed.
|
|
-- This prevents accidental conversions.
|
|
--
|
|
-- Otherwise, when the file's inode is the same as one that was used for
|
|
-- annexed content before, annex it. This handles cases such as renaming an
|
|
-- unlocked annexed file followed by git add, which the user naturally
|
|
-- expects to behave the same as git mv.
|
|
shouldAnnex :: RawFilePath -> Maybe (Sha, FileSize, ObjectType) -> Maybe Key -> Annex Bool
|
|
shouldAnnex file indexmeta moldkey = ifM (annexGitAddToAnnex <$> Annex.getGitConfig)
|
|
( checkunchangedgitfile $ checkmatcher checkheuristics
|
|
, checkunchangedgitfile checkheuristics
|
|
)
|
|
where
|
|
checkmatcher d
|
|
| dotfile file = ifM (getGitConfigVal annexDotFiles)
|
|
( go
|
|
, d
|
|
)
|
|
| otherwise = go
|
|
where
|
|
go = do
|
|
matcher <- largeFilesMatcher
|
|
checkFileMatcher' matcher file d
|
|
|
|
checkheuristics = case moldkey of
|
|
Just _ -> return True
|
|
Nothing -> checkknowninode
|
|
|
|
checkknowninode = withTSDelta (liftIO . genInodeCache file) >>= \case
|
|
Nothing -> pure False
|
|
Just ic -> Database.Keys.isInodeKnown ic =<< sentinalStatus
|
|
|
|
-- This checks for a case where the file had been added to git
|
|
-- previously, not to the annex before, and its content is not
|
|
-- changed, but git is running the clean filter again on it
|
|
-- (eg because its mtime or inode changed, or just because git feels
|
|
-- like it). Such a file should not be added to the annex, even if
|
|
-- annex.largefiles now matches it, because the content is not
|
|
-- changed.
|
|
checkunchangedgitfile cont = case (moldkey, indexmeta) of
|
|
(Nothing, Just (sha, sz, _)) -> liftIO (catchMaybeIO (getFileSize file)) >>= \case
|
|
Just sz' | sz' == sz -> do
|
|
-- The size is the same, so the file
|
|
-- is not much larger than what was stored
|
|
-- in git before, so it won't be out of
|
|
-- line to hash it. However, the content
|
|
-- is prevented from being stored in git
|
|
-- when hashing.
|
|
h <- inRepo $ hashObjectStart False
|
|
sha' <- liftIO $ hashFile h file
|
|
liftIO $ hashObjectStop h
|
|
if sha' == sha
|
|
then return False
|
|
else cont
|
|
_ -> cont
|
|
_ -> cont
|
|
|
|
emitPointer :: Key -> IO ()
|
|
emitPointer = S.putStr . formatPointer
|
|
|
|
-- Recover from a previous race between eg git mv and git-annex get.
|
|
-- That could result in the file remaining a pointer file, while
|
|
-- its content is present in the annex. Populate the pointer file.
|
|
--
|
|
-- This also handles the case where a copy of a pointer file is made,
|
|
-- then git-annex gets the content, and later git add is run on
|
|
-- the pointer copy. It will then be populated with the content.
|
|
getMoveRaceRecovery :: Key -> RawFilePath -> Annex ()
|
|
getMoveRaceRecovery k file = void $ tryNonAsync $
|
|
whenM (inAnnex k) $ do
|
|
obj <- calcRepo (gitAnnexLocation k)
|
|
-- Cannot restage because git add is running and has
|
|
-- the index locked.
|
|
populatePointerFile (Restage False) k obj file >>= \case
|
|
Nothing -> return ()
|
|
Just ic -> Database.Keys.addInodeCaches k [ic]
|
|
|
|
update :: CommandStart
|
|
update = do
|
|
updateSmudged (Restage True)
|
|
stop
|
|
|
|
updateSmudged :: Restage -> Annex ()
|
|
updateSmudged restage = streamSmudged $ \k topf -> do
|
|
f <- fromRepo (fromTopFilePath topf)
|
|
whenM (inAnnex k) $ do
|
|
obj <- calcRepo (gitAnnexLocation k)
|
|
unlessM (isJust <$> populatePointerFile restage k obj f) $
|
|
liftIO (isPointerFile f) >>= \case
|
|
Just k' | k' == k -> toplevelWarning False $
|
|
"unable to populate worktree file " ++ fromRawFilePath f
|
|
_ -> noop
|