ghc -e 'Control.Monad.Fix.mfix error' panics

[duncan]

Just for the record, a low priority bug...

Folk on IRC found this obscure torture test:

Control.Monad.Fix.mfix error

Depending on how you run it it gives different errors.

Compiling a module main = mfix error works fine:

ghc foo.hs -o foo
./foo
foo: <<loop>>

It exits with a non-zero exit code as expected.

ghci fares a little worse but ok:

Prelude> Control.Monad.Fix.mfix error
*** Exception: *** Exception: <<loop>>

So actually the only way that goes wrong is ghc -e :

ghc -e 'Control.Monad.Fix.mfix error'
*** Exception: ghc-6.4.2: panic! (the `impossible' happened, GHC version 6.4.2):
        <<loop>>

Please report this as a compiler bug.  See:
  http://www.haskell.org/ghc/reportabug

Same behaviour with recent ghc-6.5.

So I'd guess it's just the catching of errors in ghc -e that's not quite right, probably because of the way that tring to generate the error message generates an error.