[llvm] [LVI] Handle icmp of ashr. (PR #68010)

[Amara Emerson via llvm-commits]

aemerson wrote:

> This doesn't compute the correct ranges for slt and sge predicates, because the +/-1 adjustment needs to happen on C prior to shifting, not the result.
> 
> I gave my suggestion from the Phabricator review a try with a patch like this: https://gist.github.com/nikic/e040c8870eb6d4aa59057b332ff1ce40 I had to adjust some of the test expectations, but I believe that's because they weren't correct due to the issue above.

Thanks, I've merged your patch into this.

https://github.com/llvm/llvm-project/pull/68010