Avoid crashing when run in a bare git repo that somehow contains an index file

Do not populate the keys database with associated files,
because a bare repo has no working tree, and so it does not make sense to
populate it.

Queries of associated files in the keys database always return empty lists
in a bare repo, even if it's somehow populated. One way it could be
populated is if a user converts a non-bare repo to a bare repo.

Note that Git.Config.isBare does a string comparison, so this is not free!
But, that string comparison is very small compared to a sqlite query.

Sponsored-by: Erik Bjäreholt on Patreon

diff --git a/CHANGELOG b/CHANGELOG
index 16cd4f00985d0154e851966c6e4525c3d411a86d..88f7ce5a5e3b29f2a4385af7287211884bac8c8d 100644 (file)
--- a/CHANGELOG
+++ b/CHANGELOG
@@ -10,6 +10,8 @@ git-annex (8.20211232) UNRELEASED; urgency=medium
     a retry from failing again. 
     (reversion introduced in version 8.20210903)
   * adb: Added ignorefinderror configuration parameter.
+  * Avoid crashing when run in a bare git repo that somehow contains an
+    index file.
 
  -- Joey Hess <id@joeyh.name>  Mon, 03 Jan 2022 14:01:14 -0400
 

diff --git a/Database/Keys.hs b/Database/Keys.hs
index 386c3245b12b2d7d17a7e8e9c86bc33e771f1063..a342274f5ec654292e191a705ffb6961a2247626 100644 (file)
--- a/Database/Keys.hs
+++ b/Database/Keys.hs
@@ -49,6 +49,7 @@ import Git.Sha
 import Git.CatFile
 import Git.Branch (writeTreeQuiet, update')
 import qualified Git.Ref
+import qualified Git.Config
 import Config.Smudge
 import qualified Utility.RawFilePath as R
 
@@ -154,11 +155,23 @@ addAssociatedFile k f = runWriterIO $ SQL.addAssociatedFile k f
 {- Note that the files returned were once associated with the key, but
  - some of them may not be any longer. -}
 getAssociatedFiles :: Key -> Annex [TopFilePath]
-getAssociatedFiles = runReaderIO . SQL.getAssociatedFiles
+getAssociatedFiles k = emptyWhenBare $ runReaderIO $ SQL.getAssociatedFiles k
+
+{- Queries for associated files never return anything when in a bare
+ - repository, since without a work tree there can be no associated files. 
+ -
+ - Normally the keys database is not even populated with associated files 
+ - in a bare repository, but it might happen if a non-bare repo got
+ - converted to bare. -}
+emptyWhenBare :: Annex [a] -> Annex [a]
+emptyWhenBare a = ifM (Git.Config.isBare <$> gitRepo)
+       ( return []
+       , a
+       )
 
 {- Include a known associated file along with any recorded in the database. -}
 getAssociatedFilesIncluding :: AssociatedFile -> Key -> Annex [RawFilePath]
-getAssociatedFilesIncluding afile k = do
+getAssociatedFilesIncluding afile k = emptyWhenBare $ do
        g <- Annex.gitRepo
        l <- map (`fromTopFilePath` g) <$> getAssociatedFiles k
        return $ case afile of
@@ -168,7 +181,7 @@ getAssociatedFilesIncluding afile k = do
 {- Gets any keys that are on record as having a particular associated file.
  - (Should be one or none but the database doesn't enforce that.) -}
 getAssociatedKey :: TopFilePath -> Annex [Key]
-getAssociatedKey = runReaderIO . SQL.getAssociatedKey
+getAssociatedKey f = emptyWhenBare $ runReaderIO $ SQL.getAssociatedKey f
 
 removeAssociatedFile :: Key -> TopFilePath -> Annex ()
 removeAssociatedFile k = runWriterIO . SQL.removeAssociatedFile k
@@ -233,7 +246,7 @@ isInodeKnown i s = or <$> runReaderIO ((:[]) <$$> SQL.isInodeKnown i s)
  - is an associated file.
  -}
 reconcileStaged :: H.DbQueue -> Annex ()
-reconcileStaged qh = do
+reconcileStaged qh = unlessM (Git.Config.isBare <$> gitRepo) $ do
        gitindex <- inRepo currentIndexFile
        indexcache <- fromRawFilePath <$> fromRepo gitAnnexKeysDbIndexCache
        withTSDelta (liftIO . genInodeCache gitindex) >>= \case

diff --git a/doc/bugs/Repo_manipulation_breaks_git-annex_drop_--all.mdwn b/doc/bugs/Repo_manipulation_breaks_git-annex_drop_--all.mdwn
index 9cbb3926ff0999de5a6735a4d26bef7044539dec..1c861fd5aa150cfd8f6b3aff70e8ed7209dce06e 100644 (file)
--- a/doc/bugs/Repo_manipulation_breaks_git-annex_drop_--all.mdwn
+++ b/doc/bugs/Repo_manipulation_breaks_git-annex_drop_--all.mdwn
@@ -55,3 +55,4 @@ AFAICS this creates the same records as if I would have done this in a regular w
 
 Is this a bug, or am I doing something wrong? Thanks in advance for your time!
 
+> [[fixed|done]] --[[Joey]]

diff --git a/doc/bugs/Repo_manipulation_breaks_git-annex_drop_--all/comment_1_22fb9214e6e31d04b0ffe3266d26ea69._comment b/doc/bugs/Repo_manipulation_breaks_git-annex_drop_--all/comment_1_22fb9214e6e31d04b0ffe3266d26ea69._comment
new file mode 100644 (file)
index 0000000..082b7b7
--- /dev/null
+++ b/doc/bugs/Repo_manipulation_breaks_git-annex_drop_--all/comment_1_22fb9214e6e31d04b0ffe3266d26ea69._comment
@@ -0,0 +1,28 @@
+[[!comment format=mdwn
+ username="joey"
+ subject="""comment 1"""
+ date="2022-01-11T16:25:50Z"
+ content="""
+`git update-index` creates an index file. A repo with an index file is no
+longer a bare repo to a certain extent, at least it is very unusual for a
+repo to be both bare and contain an index file, since after all an index
+file is a record of the files in the working tree. This seems to be the root
+of the confusion.
+
+I was able to reproduce this by simply copying `.git/index` from a non-bare
+repo into the bare repo and then running `git-annex drop --all`.
+
+Removing the index file was not sufficient to fix it. It turned out I also 
+needed to delete `annex/keydb*`. Then things returned to normal.
+
+So, it seems that the keys database is getting populated in a bare repo
+when there's an index file, and once the keys database is populated, it
+runs code paths that will not work in a bare repo, because that database
+contains paths (taken from the index) that it treats as being present in a
+nonexistant working tree. I've fixed it to both avoid populating the keys
+database, and ignore a populated keys database in this situation.
+
+But.. My suggestion is, if you need to do this kind of thing in a bare repo,
+set `GIT_INDEX_FILE` to some other file. That's how git-annex makes similar
+tree objects.
+"""]]