In Rocq 9.1.0 with stdlib 9.0.0, rewriting with a hypothesis of type JMeq x y works out of the box. However, in Rocq 9.2+rc1 with the current master branch of stdlib, the same code fails, even after explicitly enabling rewriting for JMeq. This appears to be a regression, but it is unclear whether the issue lies in Rocq itself or in stdlib.
Working behavior (Rocq 9.1.0 + stdlib 9.0.0):
From Stdlib Require Export JMeq.
Goal forall A (x y : A), JMeq x y -> x = y.
intros A x y H.
rewrite H.
(* Goal becomes: y = y *)
Failing behavior (Rocq 9.2+rc1 + stdlib master):
From Stdlib Require Export JMeq.
Scheme Rewriting for JMeq.
Goal forall A (x y : A), JMeq x y -> x = y.
intros A x y H.
Fail rewrite H.
Despite the Scheme Rewriting for JMeq command, rewrite H is rejected.
Is this behavior change expected while stdlib is being adapted to Rocq 9.2,
or should rewriting with JMeq continue to work in this situation?