[Bug middle-end/50865] Invalid code generation for INT64_MIN % 1 on x86_64

[glisse at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=50865

--- Comment #17 from Marc Glisse <glisse at gcc dot gnu.org> ---
(In reply to Jakub Jelinek from comment #16)
> I think it doesn't.  Plus, am not sure if it would be safe to use it
> everywhere in the middle-end (plus, what about LTO, e.g. when you mix
> C/C++/... and Fortran code),

Yes, I was thinking about having the front-end add this information when it
produces the initial gimple. But value ranges go with SSA_NAMEs, and front-ends
are more familiar with variables, so I don't know how hard it would be.

> while user code should not end up with this
> values, I'd artificial compiler generated code could still trigger those.

Is there already a lot of artificial code at that stage (that the front-end
does not clearly identify as such)?

(just discussing this in passing, I have no particular interest in fortran)