Support GHC 9.6.4

[istathar]

Bump resolver to lts-22.9 which builds against GHC 9.6.4 for testing. Testing with VS Code and haskell-language-server 2.6.0.0 via GHCup.

[istathar]

@guaraqe not sure if you're still using core-effect-effectful but when I attempted to upgrade to GHC 9.6.4 and the current Stackage snapshot we got a build error,

core-effect-effectful        > /home/runner/work/unbeliever/unbeliever/core-effect-effectful/lib/Core/Effect/Effectful.hs:204:5: error: [GHC-83865]
core-effect-effectful        >     • Couldn't match expected type: Effect.Eff es α
core-effect-effectful        >                   with actual type: ((forall r. Effect.Eff es0 r -> IO r) -> IO a0)
core-effect-effectful        >                                     -> Effect.Eff es0 a0
core-effect-effectful        >     • Probable cause: ‘($)’ is applied to too few arguments
core-effect-effectful        >       In a stmt of a 'do' block:
core-effect-effectful        >         Effect.withEffToIO
core-effect-effectful        >           $ \ runInIO -> subProgram context1 $ do action $ \ inner -> ...
core-effect-effectful        >       In the expression:
core-effect-effectful        >         do ProgramE context1 <- Effect.getStaticRep
core-effect-effectful        >            Effect.withEffToIO $ \ runInIO -> subProgram context1 $ do ...
core-effect-effectful        >       In an equation for ‘withProgram’:
core-effect-effectful        >           withProgram action
core-effect-effectful        >             = do ProgramE context1 <- Effect.getStaticRep
core-effect-effectful        >                  Effect.withEffToIO $ \ runInIO -> subProgram context1 $ ...
core-effect-effectful        >     • Relevant bindings include
core-effect-effectful        >         action :: (forall β. Effect.Eff es β -> Program τ β) -> Program τ α
core-effect-effectful        >           (bound at lib/Core/Effect/Effectful.hs:198:13)
core-effect-effectful        >         withProgram :: ((forall β. Effect.Eff es β -> Program τ β)
core-effect-effectful        >                         -> Program τ α)
core-effect-effectful        >                        -> Effect.Eff es α
core-effect-effectful        >           (bound at lib/Core/Effect/Effectful.hs:198:1)
core-effect-effectful        >     |
core-effect-effectful        > 204 |     Effect.withEffToIO $ \runInIO ->
core-effect-effectful        >     |     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^...
core-effect-effectful        > 

if you have an immediate understanding of what might have broken then a fix would be welcome! Otherwise we'll just comment out the [testing] dependency for now and you can return to if you ever need to.

[istathar]

I've removed core-effect-effectful from the default build. It's not part of the core set so it can wait on the side until someone has a chance to look into this.

[istathar]

Now it's failing because of HsOpenSSL having an "incompatible pointer type"

HsOpenSSL             > /tmp/stack-3146af9db5b21892/HsOpenSSL-0.11.7.6//tmp/ghc136953_0/ghc_178.c:18:175: error:
HsOpenSSL             >      error: passing argument 4 of ‘DSA_get0_pqg’ from incompatible pointer type [-Wincompatible-pointer-types]
HsOpenSSL             >        18 | void ghczuwrapperZC3ZCHsOpenSSLzm0zi11zi7zi6zmIjQSlutIIqFJA1gCaYa4XgZCOpenSSLziDSAZCDSAzuget0zupqg(void* a1, BIGNUM** a2, BIGNUM** a3, BIGNUM** a4) {DSA_get0_pqg(a1, a2, a3, a4);}
HsOpenSSL             >           |                                                                                                                                                                               ^~
HsOpenSSL             >           |                                                                                                                                                                               |
HsOpenSSL             >           |                                                                                                                                                                               BIGNUM ** {aka struct bignum_st **}
HsOpenSSL             >    |
HsOpenSSL             > 18 | void ghczuwrapperZC3ZCHsOpenSSLzm0zi11zi7zi6zmIjQSlutIIqFJA1gCaYa4XgZCOpenSSLziDSAZCDSAzuget0zupqg(void* a1, BIGNUM** a2, BIGNUM** a3, BIGNUM** a4) {DSA_get0_pqg(a1, a2, a3, a4);}
HsOpenSSL             >    |                                                                                                                                                                               ^
HsOpenSSL             > 

what on earth do we do about that one?

[istathar]

Looks like this is the issue:

• fails to build with Fedora 40 (gcc-14) haskell-cryptography/HsOpenSSL#88

Annoyingly we have to put

ghc-options:
  HsOpenSSL: -optc=-Wno-incompatible-pointer-types

in the stack.yaml of any project that uses our libraries in order for the dependencies to build.