runTSimple fails to compile when eta-reduced

[qwbarch]

This is a pretty minor issue, but it's bothering me since I can't compose functions when using runTSimple.
Here is an example of what compiles and what doesn't:

import Polysemy (InterpreterFor, interpretH, makeSem, runTSimple)

data Foo m a where
  Bar :: (Int -> m a) -> Foo m ()

makeSem ''Foo

interpretFoo :: InterpreterFor Foo r
interpretFoo = interpretH $ \case
  Bar f -> do
    let getInt = pure 0
    -- this compiles
    getInt >>= \x -> runTSimple (f x)
    -- this fails to compile
    getInt >>= (runTSimple . f)

Any ideas on why this is happening?

[tek]

Assuming you mean runTSimple . f, this is certainly due to the temporary change introduced in GHC 9 and reverted in 9.2.4 called Simplified Subsumption. There is no way around it other than upgrading GHC.

[qwbarch]

Whoops good catch, I didn't notice the missing compose function earlier.
I'm just going to bump my ghc version to avoid the change then. Thanks a lot!

[tek]

note that HLS does not work well with Polysemy on GHC 9.2 (issue)

[qwbarch]

Hmm that's unfortunate. I'll try downgrading to 8.10.7 for now then. Thanks for pointing that out

[tek]

I think it might be easier to just eta-expand 🙂

[qwbarch]

My code-base should be compatible with 8.10.7, so I might as well take the extra step to get rid of the ugliness of eta-expansions hehe