Disable git-annex's support for GIT_SSH and GIT_SSH_COMMAND, unless GIT_ANNEX_USE_GIT_SSH=1 is also set in the environment.

This is necessary because as feared, the extra -n parameter that git-annex
passes breaks uses of these environment variables that expect exactly the
parameters that git passes.

For example, see https://github.com/datalad/datalad/issues/1456

It would of course be possible to pre-close stdin before running ssh so not
needing the -n, and I think that would not even break ssh's password
caching. But it would probably involve a lot of work, possibly would need
to deal with some layering violations, and would be error-prone. The really
clean fix would be to make all the ssh stuff return a CreateProcess, which
could have the handle closed when appropriate, but that would be a large
reworing of the code base.

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

Annex/Ssh.hs

@@ -52,10 +52,13 @@ data ConsumeStdin = ConsumeStdin | NoConsumeStdin
 
 {- Generates a command to ssh to a given host (or user@host) on a given
  - port. This includes connection caching parameters, and any ssh-options.
- - If GIT_SSH or GIT_SSH_COMMAND is set, they are used instead. -}
+ - If GIT_SSH or GIT_SSH_COMMAND is enabled, they are used instead. -}
 sshCommand :: ConsumeStdin -> (SshHost, Maybe SshPort) -> RemoteGitConfig -> SshCommand -> Annex (FilePath, [CommandParam])
-sshCommand cs (host, port) gc remotecmd = maybe go return
-	=<< liftIO (gitSsh' host port remotecmd (consumeStdinParams cs))
+sshCommand cs (host, port) gc remotecmd = ifM (liftIO safe_GIT_SSH)
+	( maybe go return
+		=<< liftIO (gitSsh' host port remotecmd (consumeStdinParams cs))
+	, go
+	)
   where
 	go = do
 		ps <- sshOptions cs (host, port) gc []
@@ -81,6 +84,12 @@ sshOptions cs (host, port) gc opts = go =<< sshCachingInfo (host, port)
 		, [Param "-T"]
 		]
 
+{- Due to passing -n to GIT_SSH and GIT_SSH_COMMAND, some settings
+ - of those that expect exactly git's parameters will break. So only
+ - use those if the user set GIT_ANNEX_USE_GIT_SSH to say it's ok. -}
+safe_GIT_SSH :: IO Bool
+safe_GIT_SSH = (== Just "1") <$> getEnv "GIT_ANNEX_USE_GIT_SSH"
+
 consumeStdinParams :: ConsumeStdin -> [CommandParam]
 consumeStdinParams ConsumeStdin = []
 consumeStdinParams NoConsumeStdin = [Param "-n"]
@@ -305,13 +314,13 @@ inRepoWithSshOptionsTo remote gc a =
  - to set GIT_SSH=git-annex, and set sshOptionsEnv when running git
  - commands.
  -
- - If GIT_SSH or GIT_SSH_COMMAND are set, this has no effect. -}
+ - If GIT_SSH or GIT_SSH_COMMAND are enabled, this has no effect. -}
 sshOptionsTo :: Git.Repo -> RemoteGitConfig -> Git.Repo -> Annex Git.Repo
 sshOptionsTo remote gc localr
 	| not (Git.repoIsUrl remote) || Git.repoIsHttp remote = unchanged
 	| otherwise = case Git.Url.hostuser remote of
 		Nothing -> unchanged
-		Just host -> ifM (liftIO gitSshEnvSet)
+		Just host -> ifM (liftIO $ safe_GIT_SSH <&&> gitSshEnvSet)
 			( unchanged
 			, do
 				(msockfile, _) <- sshCachingInfo (host, Git.Url.port remote)

CHANGELOG

@@ -7,6 +7,11 @@ git-annex (6.20170322) UNRELEASED; urgency=medium
   * Added remote.<name>.annex-push and remote.<name>.annex-pull
     which can be useful to make remotes that don't get fully synced with
     local changes.
+  * Disable git-annex's support for GIT_SSH and GIT_SSH_COMMAND, unless
+    GIT_ANNEX_USE_GIT_SSH=1 is also set in the environment. This is
+    necessary because as feared, the extra -n parameter that git-annex
+    passes breaks uses of these environment variables that expect exactly
+    the parameters that git passes.
 
  -- Joey Hess <id@joeyh.name>  Wed, 29 Mar 2017 12:41:46 -0400
 

doc/git-annex.mdwn

@@ -1434,7 +1434,10 @@ These environment variables are used by git-annex when set:
   Handled similarly to the same as described in git(1).
   The one difference is that git-annex will sometimes pass an additional
   "-n" parameter to these, as the first parameter, to prevent ssh from
-  reading from stdin.
+  reading from stdin. Since that can break existing uses of these
+  environment variables that don't expect the extra parameter, you will
+  need to set `GIT_ANNEX_USE_GIT_SSH=1` to make git-annex support
+  these.
 
   Note that setting either of these environment variables prevents
   git-annex from automatically enabling ssh connection caching