[Bug tree-optimization/71661] [7 Regression] wrong code at -O3

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71661

--- Comment #7 from Richard Biener <rguenth at gcc dot gnu.org> ---
(In reply to Jeffrey A. Law from comment #5)
> We can likely invalidate the cached information, but ISTM that there's a
> reasonable chance we're going to be playing whack-a-mole with this stuff.

Unfortunately playing whack-a-mole is what we need to do here, as with any
other on-the-side information we keep up-to-date (like range info, points-to,
EH info, etc.).