I am trying to work out empirically whether empty statement lists are allowed in a stmtlist, and one of the points these are used in Parser.y is for recursive do.

If I try to load the following

----------------
{-# LANGUAGE RecursiveDo #-}
bar :: IO ()
bar = do
rec {}
return ()
-----------------------

into GHCI, I get a "Prelude.head: empty list" panic.

I think this is related to `rnRecStmtsAndThen`

If I try

---------------------
bar :: IO ()
bar = do

do {}

return ()
-----------------------

I get a warning about empty do statements not being allowed.

Hence the question as to wether recursive do should be allowed to have an empty statement list.

I am trying to manage the API annotations for semicolons in statement lists, and the current way it is being parsed makes this difficult, I would prefer to simplify it to something like how import decls are parsed, which can also have multiple semicolons in between.

Alan