From f2c08124b30eb87482dc0ed1d7199aa58950e309 Mon Sep 17 00:00:00 2001
From: Bodigrim <andrew.lelechenko@gmail.com>
Date: Fri, 22 Apr 2022 19:03:04 +0100
Subject: [PATCH] Document behaviour of RULES with KnownNat
---
docs/users_guide/exts/rewrite_rules.rst | 9 +++++++++
1 file changed, 9 insertions(+)
diff --git a/docs/users_guide/exts/rewrite_rules.rst b/docs/users_guide/exts/rewrite_rules.rst
index 7ac93bb21df9..75ee2474a984 100644
--- a/docs/users_guide/exts/rewrite_rules.rst
+++ b/docs/users_guide/exts/rewrite_rules.rst
@@ -228,6 +228,15 @@ From a semantic point of view:
because ``y`` can match against ``0``.
+- A rule that has a forall binder with a polymorphic type, is likely to fail to fire. E. g., ::
+
+ {-# RULES forall (x :: forall a. Num a => a -> a). f x = blah #-}
+
+ Here ``x`` has a polymorphic type. This applies to a forall'd binder with a type class constraint, such as::
+
+ {-# RULES forall @m (x :: KnownNat m => Proxy m). g x = blah #-}
+
+ See `#21093 <https://gitlab.haskell.org/ghc/ghc/-/issues/21093>`_ for discussion.
.. _rules-inline:
--
GitLab