On Tue, Nov 18, 2014 at 11:35:50AM +0100, Richard Henderson wrote:
> This has in effect already been applied.  Note that FFI_TYPE_VOID now appears
> in the switch statement.  The abort now only applies to true out-of-bounds, aka
> corrupt, ffi_type data.

Ah, I see.  The first half of the patch still looks useful though.
Updated patch attached.

Ciao

Dominik ^_^  ^_^

-- 

Dominik Vogt
IBM Germany