[Bug tree-optimization/113210] [14 Regression] ICE: tree check: expected integer_cst, have cond_expr in get_len, at tree.h:6481

[rguenth at gcc dot gnu.org via Gcc-bugs]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113210

--- Comment #9 from Richard Biener <rguenth at gcc dot gnu.org> ---
(In reply to Jakub Jelinek from comment #8)
> Created attachment 56993 [details]
> gcc14-pr113210.patch
> 
> So, I'd go with this patch (so far untested).

I think we want to keep the invariant that both are INTEGER_CST when one is.

If we can fold the add to 1 why can't we fold the original to 0?

That is, another fix might be to adjust NITERSM1 to NITERS - 1 when
NITERS went constant ...  (btw, I want to get rid of _NITERS and only
keep _NITERSM1 given _NITERS cannot be represented when the number of
latch iterations is TYPE_MAX_VALUE)