[Bug middle-end/103483] context-sensitive ranges change triggers stringop-overread

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103483

--- Comment #9 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
(In reply to Andrew Pinski from comment #8)
> I think we should come up with a better plan in general for "flow sensative"
> warnings really. Maybe only enable them with -O2 and above. But we keep on
> getting more and more of them where it is only defined at runtime if it is
> hit.
> The other thing is not having it in -W -Wall. Or maybe even adding a way
> (outside of -fsantizer=*) to have runtime checks inside the flow where this
> happen.

Just to note, I was wrong before in talking about this case (and others)
because I didn't realize how much code is going to run into these issues. I
think Aldy did put it correct when he wrote: "These false positives "by design"
arguments are just a cop-out.".