Should `INT_MIN % -1` return 0 instead of panicking?

[glaebhoerl]

As @bill-myers originally pointed out here INT_MIN % -1 is mathematically well-defined to be 0, but Rust currently unconditionally panics instead, as like INT_MIN / -1 it is technically an overflow according to the two's complement representation. If we're incurring a branch to test for this case anyways, then we might as well return the correct result.

("Forked" from #1237.)

[Emerentius]

Ran into an unexpected panic with the gcd function in the num crate for this reason (euclidean algorithm). The error message telling me that there was an overflow wasn't very helpful either because I just didn't expect it to be from this.

[ticki]

👍 We should generally avoid panics.

[dgrunwald]

👍 0 is obviously the correct result.

C# also handles this the same way: int.MinValue % -1 is 0; int.MinValue / -1 is an overflow error.

[comex]

👍