[lto] implement awful types_compatible_p hack

[Kenneth Zadeck]

Toon Moene wrote:
> Nathan Froyd wrote:
>
>> 253.perlbmk now compiles (with -fwrapv) and runs properly.  This leaves
>> a miscompile of 176.gcc as the only remaining C SPECint benchmark that
>> LTO doesn't work with.
>
> Ahhhhhh ... Does that mean I can start writing my GCC Summit 2008
> paper: "Link Time Optimization: What's in it for Fortran ?"
>
> Thanks for all the great work you're doing on this branch !
>
unfortunately no

we only do c at this point.

and just between you and me, none of this is going to really help
fortran unless you can get real high level array operations into gimple
and we solve carrying the language specific aliasing info into lto.


kenny