[Bug tree-optimization/125501] [14/15/16/17 Regression] wrong code at -O3 that crashes at runtime

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=125501

--- Comment #13 from Drea Pinski <pinskia at gcc dot gnu.org> ---
(In reply to Andrew Macleod from comment #12)
> So either this isn't safe, or it is unsafe for DOM to set a global value
> based on all available ranges like it does.. which could be the case..  the
> __builtin_unreachable code in VRP went through some ppin to determine when
> it can and when it cant set those kinds of ranges.  I notice VRP1 didnt set
> the range in this case.   Perhaps DOM needs to be tightened up, or just
> leave it to ranger to do this with unreachables.


After seeing this, I get the feeling this is related to the failure that I am
getting with gcc.dg/pr80776-1.c (in the
https://gcc.gnu.org/pipermail/gcc-patches/2026-June/718954.html thread). That
is how DOM handles unreachable is getting in the way in my case though not
producing wrong code in pr80776-1.c case.