zyla
commented
2 years ago I can reproduce the error with the following program, on both GHC 9.2.4 and 9.4.2:
{-# LANGUAGE DeriveGeneric #-}
{-# OPTIONS_GHC -fconstraint-solver-iterations=4 #-}
module EnumTest where
import GHC.Generics
import Test.QuickCheck
import Test.QuickCheck.Arbitrary.Generic
data Country = PL | GB | RU | RO | CZ | HR | SK | DE | NL | ES | BR
deriving (Generic)
instance Arbitrary Country where
arbitrary = genericArbitrary
shrink = genericShrinkBehavior is different depending on -fconstraint-solver-iterations:
-
with
-fconstraint-solver-iterations=4, I get bothsolveWanteds: too many iterations (limit = 4)andAmbiguous type variables ‘af0’, ‘bf0’ arising from a use of ‘genericArbitrary’See full error message
``` /workspace/generic-arbitrary/test/EnumTest.hs:1:1: error: solveWanteds: too many iterations (limit = 4) Unsolved: WC {wc_simple = [WD] $dFiniteSum_a6Zf {7}:: FiniteSum Country (M1 C ('MetaCons "RO" 'PrefixI 'False) U1) (M1 C ('MetaCons "CZ" 'PrefixI 'False) U1) af0 bf0 (CDictCan(psc)) [WD] $dFiniteSum_a6Zv {7}:: FiniteSum Country (M1 C ('MetaCons "SK" 'PrefixI 'False) U1) (M1 C ('MetaCons "DE" 'PrefixI 'False) U1) af0 bf0 (CDictCan(psc)) [WD] $dFiniteSum_a6ZL {7}:: FiniteSum Country (M1 C ('MetaCons "ES" 'PrefixI 'False) U1) (M1 C ('MetaCons "BR" 'PrefixI 'False) U1) af0 bf0 (CDictCan(psc))} Set limit with -fconstraint-solver-iterations=n; n=0 for no limit | 1 | {-# LANGUAGE DeriveGeneric #-} | ^ /workspace/generic-arbitrary/test/EnumTest.hs:14:15: error: • Ambiguous type variables ‘af0’, ‘bf0’ arising from a use of ‘genericArbitrary’ prevents the constraint ‘(FiniteSum Country (M1 C ('MetaCons "RO" 'PrefixI 'False) U1) (M1 C ('MetaCons "CZ" 'PrefixI 'False) U1) af0 bf0)’ from being solved. Probable fix: use a type annotation to specify what ‘af0’, ‘bf0’ should be. These potential instances exist: instance (FiniteSumElem self b, Finite self a ~ 'False, Finite self b ~ 'True) => FiniteSum self a b 'False 'True -- Defined in ‘Test.QuickCheck.Arbitrary.Generic’ instance (FiniteSumElem self a, Finite self a ~ 'True, Finite self b ~ 'False) => FiniteSum self a b 'True 'False -- Defined in ‘Test.QuickCheck.Arbitrary.Generic’ instance (FiniteSumElem self a, FiniteSumElem self b, Finite self a ~ 'True, Finite self b ~ 'True) => FiniteSum self a b 'True 'True -- Defined in ‘Test.QuickCheck.Arbitrary.Generic’ • In the expression: genericArbitrary In an equation for ‘arbitrary’: arbitrary = genericArbitrary In the instance declaration for ‘Arbitrary Country’ | 14 | arbitrary = genericArbitrary | ^^^^^^^^^^^^^^^^ ``` -
with
-fconstraint-solver-iterations=5, I see onlysolveWanteds: too many iterations (limit = 5), no ambiguous type errorSee full error message
``` /workspace/generic-arbitrary/test/EnumTest.hs:1:1: error: solveWanteds: too many iterations (limit = 5) Unsolved: WC {} Set limit with -fconstraint-solver-iterations=n; n=0 for no limit | 1 | {-# LANGUAGE DeriveGeneric #-} | ^ ``` -
with
-fconstraint-solver-iterations=6, it compiles fine.
I don't know much about the GHC constraint solver, but maybe the ambiguous type error can arise due to an intermediate state the solver is left in after iteration limit is exhausted?
s9gf4ult
arybczak
With a plain sum type, on GHC 9.4.1, we get ambiguous type variable errors when trying to write
instance Arbitrary X where arbitrary = genericArbitraryApologies for the bad issue, I'll work up a reproduction soon