Can't compile with GHC 9.2.1

[LeventErkok]

I can't quite tell if this is a problem with hlint itself or maybe the dependencies went out of sync, but I can no longer compile hlint with GHC 9.2.1:

Building library for hlint-3.3.4..
[ 1 of 59] Compiling EmbedData        ( src/EmbedData.hs, dist/build/EmbedData.o, dist/build/EmbedData.dyn_o )
[ 2 of 59] Compiling Extension        ( src/Extension.hs, dist/build/Extension.o, dist/build/Extension.dyn_o )

src/Extension.hs:64:25: error:
    • Couldn't match type ‘ghc-boot-th-9.2.1:GHC.LanguageExtensions.Type.Extension’
                     with ‘Extension’
      Expected: [(Extension, ([Extension], [Extension]))]
        Actual: [(ghc-boot-th-9.2.1:GHC.LanguageExtensions.Type.Extension,
                  ([ghc-boot-th-9.2.1:GHC.LanguageExtensions.Type.Extension],
                   [ghc-boot-th-9.2.1:GHC.LanguageExtensions.Type.Extension]))]
      NB: ‘Extension’
            is defined in ‘GHC.LanguageExtensions.Type’
                in package ‘ghc-lib-parser-9.2.1.20211030’
          ‘ghc-boot-th-9.2.1:GHC.LanguageExtensions.Type.Extension’
            is defined in ‘GHC.LanguageExtensions.Type’
                in package ‘ghc-boot-th-9.2.1’
    • In the expression: GhclibParserEx.extensionImplications
      In an equation for ‘extensionImplications’:
          extensionImplications = GhclibParserEx.extensionImplications
   |
64 | extensionImplications = GhclibParserEx.extensionImplications
   |                         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

[shayne-fletcher]

hlint does not currently support building with ghc-9.2.1. work is underway to fix that.

[garyamorris]

might be a silly question, new to Haskell. How do I install hlint if I have 9.2.1 installed? cabal install hlint throws errors for me.

[LeventErkok]

@garyamorris You can't, until they make a new release. But as I understand it, it should be out any day now.

[LeventErkok]

@ndmitchell I just tried to compile HLint 3.3.5, which I thought was going to allow for compilation with GHC 9.2.1; but I'm still getting the exact same compilation error above. Was there something amiss?

[ndmitchell]

@LeventErkok - I made a release to allow compiling with aeson-2.0. I've merged a patch that means HEAD should compile with GHC 9.2, but the apply-refact binary (which the CI tests use) won't yet - see mpickering/apply-refact#114. Once that gets fixed I'll do a release.

[LeventErkok]

@ndmitchell @shayne-fletcher

Looks like the dependence is only for the apply-refact part. I do not use hlint to apply the suggestions automatically, and I suspect most end-users like myself do not either.

I'm wondering if you can reduce this dependency by making the "apply" feature conditional, via a cabal-flag. This would allow the mainline users to move on with GHC 9.2.1; and people who want the apply feature can stay with the old version. (Or complain louder!) Or perhaps there's some other trick to avoid this issue?

I understand this is extra work for you, both now and also for future maintenance; but hlint is an important part of the eco-system. Looks like the first commit to #1247 was on May of last year! almost 8 months ago.

Sorry for being a pest about this; just wondering if the ecosystem can be improved by reducing this dependency to some extent.

[ndmitchell]

Do @alanz or @zliu41 have any info on apply-refact? Splitting apply into pieces makes it harder for HLint maintainers, and also harder for downstream consumers like Haskell Language Server, so would be an option I'd rather not take. Depends how far away apply-refact is.

[alanz]

I am down to two failures with GHC 9.2.1, almost there.

[alanz]

mpickering/apply-refact#116 should do the trick.

[LeventErkok]

@ndmitchell @shayne-fletcher Pinging again! Looks like apply-refact is good to go?

[ndmitchell]

@LeventErkok - looks like mpickering/apply-refact#116 is not yet merged. @alanz - is that a pre-requisite? Anything we can do to move that PR forward?

[LeventErkok]

@ndmitchell I believe the merge has now gone thru. Right @zliu41 ?

[zliu41]

@LeventErkok there are currently two HLint test failures using apply-refact HEAD, and I'm looking into it.