CarpHask fails to build with hashable >= 1.4 (GHC 9.2 / Stackage LTS-20)

[sternenseemann]

Building library for CarpHask-0.5.5.0..
[ 1 of 53] Compiling Map              ( src/Map.hs, dist/build/Map.o, dist/build/Map.dyn_o )
[ 2 of 53] Compiling Set              ( src/Set.hs, dist/build/Set.o, dist/build/Set.dyn_o )
[ 3 of 53] Compiling Util             ( src/Util.hs, dist/build/Util.o, dist/build/Util.dyn_o )
[ 4 of 53] Compiling SymPath          ( src/SymPath.hs, dist/build/SymPath.o, dist/build/SymPath.dyn_o )
[ 5 of 53] Compiling Types            ( src/Types.hs, dist/build/Types.o, dist/build/Types.dyn_o )
[ 6 of 53] Compiling TypesToC         ( src/TypesToC.hs, dist/build/TypesToC.o, dist/build/TypesToC.dyn_o )
[ 7 of 53] Compiling TypePredicates   ( src/TypePredicates.hs, dist/build/TypePredicates.o, dist/build/TypePredicates.dyn_o )
[ 8 of 53] Compiling Path             ( src/Path.hs, dist/build/Path.o, dist/build/Path.dyn_o )
[ 9 of 53] Compiling Info             ( src/Info.hs, dist/build/Info.o, dist/build/Info.dyn_o )
[10 of 53] Compiling Project          ( src/Project.hs, dist/build/Project.o, dist/build/Project.dyn_o )
[11 of 53] Compiling Obj              ( src/Obj.hs, dist/build/Obj.o, dist/build/Obj.dyn_o )

src/Obj.hs:1011:10: error:
    • No instance for (Eq Context)
        arising from the superclasses of an instance declaration
    • In the instance declaration for ‘Hashable Context’
     |
1011 | instance Hashable Context where
     |          ^^^^^^^^^^^^^^^^
[35 of 53] Compiling ColorText        ( src/ColorText.hs, dist/build/ColorText.o, dist/build/ColorText.dyn_o )

The cause is probably hashable-1.4.0.0 which changed the following:

Eq is now a superclass of Hashable. Also Eq1 is a superclass of Hashable1 and Eq2 is a superclass of Hashable2 when exists.

[eriksvedang]

Thanks for reporting! Do you have a suggestion for what the correct fix is?

[scolsen]

Looks like we'd need to declare an instance Eq Context or just derive Eq

[scolsen]

So I fixed that, but there are further issues. We have a lot of "uni patterns" in let bindings let (FooData x y) = which we don't permit because we compile w/ exhaustive pattern checking, so we'll have to fix all those too. I'm guessing the prior version of GHC we were using didn't apply this check to let bindings or something.

[scolsen]

see #1449 for a fix. let's see if the tests pass, though. it required touching a lot of code