(* This requires cumulativity *) Definition Type2 := Type. Definition Type1 : Type2 := Type. Lemma lem1 : (True -> Type1) -> Type2. intro H. apply H. exact I. Qed. Lemma lem2 : forall (A : Type) (P : A -> Type) (x : A), (forall y : A, x = y -> P y) -> P x. auto. Qed. Lemma lem3 : forall P : Prop, P. intro P; pattern P. apply lem2. Abort. (* Check managing of universe constraints in inversion *) (* Bug report #855 *) Inductive dep_eq : forall X : Type, X -> X -> Prop := | intro_eq : forall (X : Type) (f : X), dep_eq X f f | intro_feq : forall (A : Type) (B : A -> Type), let T := forall x : A, B x in forall (f g : T) (x : A), dep_eq (B x) (f x) (g x) -> dep_eq T f g. Require Import Relations. Theorem dep_eq_trans : forall X : Type, transitive X (dep_eq X). Proof. unfold transitive. intros X f g h H1 H2. inversion H1. Abort. (* Submitted by Bas Spitters (bug report #935) *) (* This is a problem with the status of the type in LetIn: is it a user-provided one or an inferred one? At the current time, the kernel type-check the type in LetIn, which means that it must be considered as user-provided when calling the kernel. However, in practice it is inferred so that a universe refresh is needed to set its status as "user-provided". Especially, universe refreshing was not done for "set/pose" *) Lemma ind_unsec : forall Q : nat -> Type, True. intro. set (C := forall m, Q m -> Q m). exact I. Qed. (* Submitted by Danko Ilik (bug report #1507); related to LetIn *) Record U : Type := { A:=Type; a:A }.