From 4645ffb130eef6466a978f6168528ed9312f3a33 Mon Sep 17 00:00:00 2001 From: Richard Eisenberg Date: Wed, 16 Dec 2020 15:05:15 -0500 Subject: Cite "Kind Inference for Datatypes" --- compiler/GHC/Tc/Utils/TcMType.hs | 3 +++ 1 file changed, 3 insertions(+) diff --git a/compiler/GHC/Tc/Utils/TcMType.hs b/compiler/GHC/Tc/Utils/TcMType.hs index ccb9152e01..4d2bebe189 100644 --- a/compiler/GHC/Tc/Utils/TcMType.hs +++ b/compiler/GHC/Tc/Utils/TcMType.hs @@ -1230,6 +1230,9 @@ We do this eager erroring in candidateQTyVars, which always precedes generalisation, because at that moment we have a clear picture of what skolems are in scope within the type itself (e.g. that 'forall arg'). +This change is inspired by and described in Section 7.2 of "Kind Inference +for Datatypes", POPL'20. + Wrinkle: We must make absolutely sure that alpha indeed is not -- cgit v1.2.1