[Bug middle-end/68785] [6 Regression] valgrind reports issues with folding on x86_64

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=68785

--- Comment #11 from Richard Biener <rguenth at gcc dot gnu.org> ---
(In reply to Jakub Jelinek from comment #10)
> Created attachment 36971 [details]
> gcc6-pr68785.patch
> 
> Sure, it is undefined behavior, but I believe we should spend the minimum
> amount of code on it.  And that IMHO is this, using the APIs the way they
> were meant to be used when added.  It won't be folded, but why should we
> spend time folding something that is invalid.

Well, fine with me.  Though we already spent the time encoding parts of the
initializer - we can as well avoid repeatedly doing that just to throw
away the result (fre1, fre2, pre at least).

So IMHO the minimum time argument is moot.