[Bug middle-end/67222] [4.9/5/6 Regression] ICE in gimple_call_arg with bogus posix_memalign

[mpolacek at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=67222

--- Comment #3 from Marek Polacek <mpolacek at gcc dot gnu.org> ---
Yeah, but that for some reason doesn't ICE.