We used to do this some operations (Z.add, Z.mul), but not for others (Z.sub, Z.of_nat); this MR blocks simpl consistently for all operations involving Z that I could think of.
The rationale for blocking everything is that simpl produces very bad results, e.g. try:
Require Import ZArith.
Eval simpl in (fun n y => Z.of_nat (S n) + y)%Z.
(*
Yields:
fun (n : nat) (y : Z) =>
match y with
| 0%Z => Z.pos (Pos.of_succ_nat n)
| Z.pos y' => Z.pos (Pos.of_succ_nat n + y')
| Z.neg y' => Z.pos_sub (Pos.of_succ_nat n) y'
end
*)Note that blocking simpl for Z.of_nat used to break lia and omega in the past, see https://github.com/coq/coq/issues/5039. For lia this has been fixed, but for omega not yet. However, in !37 (merged) we decided to use lia everywhere in both std++ and Iris because it's faster, and moreover, I believe the Coq devs intend to deprecate omega in favor of lia in the future. So, I don't think we should care about omega anymore.
I also updated the item "Side-effects" in the README to document that we block simpl on Z operations.
@jjourdan @jung Any objections to this MR?
I have fixes for Iris, iris-examples, Iron, lambdaRust (which are mostly a couple of one-liners). I thus would like to merge this ASAP (so I don't get merge conflicts on my fixes).