From 1311a2bf08ac1deb16f0b3064bc1164d75858a97 Mon Sep 17 00:00:00 2001 From: Gaëtan Gilbert Date: Wed, 13 Jun 2018 14:16:49 +0200 Subject: Remove hack skipping comparison of algebraic universes in subtyping. When inferring [u <= v+k] I replaced the exception and instead add [u <= v]. This is trivially sound and it doesn't seem possible to have the one without the other (except specially for [Set <= v+k] which was already handled). I don't know an example where this used to fail and now succeeds (the point was to remove an anomaly, but the example ~~~ Module Type SG. Definition DG := Type. End SG. Module MG : SG. Definition DG := Type : Type. Fail End MG. ~~~ now fails with universe inconsistency. Fix #7695 (soundness bug!). --- kernel/modops.mli | 1 - 1 file changed, 1 deletion(-) (limited to 'kernel/modops.mli') diff --git a/kernel/modops.mli b/kernel/modops.mli index ac76d28cf..8e7e618fc 100644 --- a/kernel/modops.mli +++ b/kernel/modops.mli @@ -106,7 +106,6 @@ type signature_mismatch_error = | RecordFieldExpected of bool | RecordProjectionsExpected of Name.t list | NotEqualInductiveAliases - | NoTypeConstraintExpected | IncompatibleInstances | IncompatibleUniverses of Univ.univ_inconsistency | IncompatiblePolymorphism of env * types * types -- cgit v1.2.3