Support GHC 9.10

[TristanCacqueray]

No description provided.

[TristanCacqueray]

This turned out to be more complicated than anticipated, this change only fixes the top-level error, now the build presently fails with:

src/Proto3/Suite/DotProto/Generate/Syntax.hs:295:21: error: [GHC-83865]
    • Couldn't match expected type: HsWildCardBndrs
                                      pass0 (GenLocated SrcSpanAnnA (GHC.HsType GhcPs))
                                    -> HsExpr GhcPs
                  with actual type: HsExpr GhcPs
    • The function ‘HsAppType’ is applied to four visible arguments,
        but its type ‘XAppTypeE p
                      -> LHsExpr p -> LHsWcType (NoGhcTc p) -> HsExpr p’
        has only three
      In the first argument of ‘noLocA’, namely
        ‘(HsAppType synDef f synDef (HsWC NoExtField (parenTy t)))’
      In the expression:
        noLocA (HsAppType synDef f synDef (HsWC NoExtField (parenTy t)))
    |
295 | appAt f t = noLocA (HsAppType synDef f
    |                     ^^^^^^^^^^^^^^^^^^...

src/Proto3/Suite/DotProto/Generate/Syntax.hs:435:11: error: [GHC-83865]
    • Couldn't match type: Maybe (ExportDoc pass2) -> IE pass2
                     with: IE GhcPs
      Expected: HsName -> HsImportSpec
        Actual: GenLocated SrcSpanAnnN GHC.Types.Name.Reader.RdrName
                -> GenLocated SrcSpanAnnA (Maybe (ExportDoc pass2) -> IE pass2)
    • In the expression: noLocA . IEVar synDef . noLocA . IEName synDef
      In an equation for ‘ieName_’:
          ieName_ = noLocA . IEVar synDef . noLocA . IEName synDef
    |
435 | ieName_ = noLocA . IEVar synDef . noLocA . IEName
    |           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^...

src/Proto3/Suite/DotProto/Generate/Syntax.hs:441:14: error: [GHC-83865]
    • Couldn't match type: Maybe (ExportDoc pass1) -> IE pass1
                     with: IE GhcPs
      Expected: HsName -> HsImportSpec
        Actual: GenLocated SrcSpanAnnN GHC.Types.Name.Reader.RdrName
                -> GenLocated SrcSpanAnnA (Maybe (ExportDoc pass1) -> IE pass1)
    • In the expression:
        noLocA . IEThingAll synDef . noLocA . IEName synDef
      In an equation for ‘ieNameAll_’:
          ieNameAll_ = noLocA . IEThingAll synDef . noLocA . IEName synDef
    |
441 | ieNameAll_ = noLocA . IEThingAll synDef . noLocA . IEName
    |              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^...

src/Proto3/Suite/DotProto/Generate/Syntax.hs:646:17: error: [GHC-83865]
    • Couldn't match expected type ‘Origin’
                  with actual type ‘DoPmc -> Origin’
    • Probable cause: ‘Generated’ is applied to too few arguments
      In the expression: Generated DoPmc
      In an equation for ‘generated’: generated = Generated DoPmc
      In an equation for ‘functionLike_’:
          functionLike_ strictness name alts
            = noLocA $ mkFunBind generated name (map match alts)
            where
                generated :: Origin
                generated = Generated DoPmc
                match :: ([HsPat], HsExp) -> HsMatch
                match (pats, rhs) = mkSimpleMatch ctxt pats rhs
                ....
    |
646 |     generated = Generated
    |                 ^^^^^^^^^...

src/Proto3/Suite/DotProto/Generate/Syntax.hs:648:27: error: [GHC-83865]
    • Couldn't match expected type ‘GHC.Types.Basic.GenReason’
                  with actual type ‘DoPmc’
    • In the first argument of ‘Generated’, namely ‘DoPmc’
      In the expression: Generated DoPmc
      In an equation for ‘generated’: generated = Generated DoPmc
    |
648 |                           DoPmc
    |                           ^^^^^

src/Proto3/Suite/DotProto/Generate/Syntax.hs:652:39: error: [GHC-83865]
    • Couldn't match type: GhcPass Parsed
                     with: GenLocated SrcSpanAnnN GHC.Types.Name.Reader.RdrName
      Expected: HsMatchContext (LIdP (NoGhcTc (GhcPass Parsed)))
        Actual: HsMatchContext GhcPs
    • In the first argument of ‘mkSimpleMatch’, namely ‘ctxt’
      In the expression: mkSimpleMatch ctxt pats rhs
      In an equation for ‘match’:
          match (pats, rhs) = mkSimpleMatch ctxt pats rhs
    |
652 |     match (pats, rhs) = mkSimpleMatch ctxt pats rhs
    |                                       ^^^^

src/Proto3/Suite/DotProto/Generate/Syntax.hs:656:18: error: [GHC-83865]
    • Couldn't match type: GenLocated
                             SrcSpanAnnN GHC.Types.Name.Reader.RdrName
                     with: GhcPass Parsed
      Expected: GhcPs
        Actual: HsName
    • In the ‘mc_fun’ field of a record
      In the expression:
        FunRhs
          {mc_fun = name, mc_fixity = Prefix, mc_strictness = strictness}
      In an equation for ‘ctxt’:
          ctxt
            = FunRhs
                {mc_fun = name, mc_fixity = Prefix, mc_strictness = strictness}
    |
656 |       { mc_fun = name
    |                  ^^^^

src/Proto3/Suite/DotProto/Generate/Syntax.hs:853:17: error: [GHC-83865]
    • Couldn't match expected type ‘Origin’
                  with actual type ‘DoPmc -> Origin’
    • Probable cause: ‘Generated’ is applied to too few arguments
      In the expression: Generated DoPmc
      In an equation for ‘generated’: generated = Generated DoPmc
      In an equation for ‘case_’:
          case_ e
            = noLocA . HsCase synDef e . mkMatchGroup generated . noLocA
            where
                generated :: Origin
                generated = Generated DoPmc
    |
853 |     generated = Generated
    |                 ^^^^^^^^^...

src/Proto3/Suite/DotProto/Generate/Syntax.hs:855:27: error: [GHC-83865]
    • Couldn't match expected type ‘GHC.Types.Basic.GenReason’
                  with actual type ‘DoPmc’
    • In the first argument of ‘Generated’, namely ‘DoPmc’
      In the expression: Generated DoPmc
      In an equation for ‘generated’: generated = Generated DoPmc
    |
855 |                           DoPmc
    |                           ^^^^^

src/Proto3/Suite/DotProto/Generate/Syntax.hs:862:14: error: [GHC-83865]
    • Couldn't match expected type: t0
                                    -> GenLocated SrcSpanAnnA (HsExpr GhcPs) -> HsExpr GhcPs
                  with actual type: HsExpr p0
    • The function ‘HsLet’ is applied to five visible arguments,
        but its type ‘XLet p -> HsLocalBinds p -> LHsExpr p -> HsExpr p’
        has only three
      In the second argument of ‘($)’, namely
        ‘HsLet synDef synDef binds synDef e’
      In the expression: noLocA $ HsLet synDef synDef binds synDef e
    |
862 |     noLocA $ HsLet synDef synDef binds synDef e
    |              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

[TristanCacqueray]

On second thought, most of the errors weren't that bad, we are down to the following:

src/Proto3/Suite/DotProto/Generate/Syntax.hs:669:39: error: [GHC-83865]
    • Couldn't match type: GhcPass Parsed
                     with: GenLocated SrcSpanAnnN GHC.Types.Name.Reader.RdrName
      Expected: HsMatchContext (LIdP (NoGhcTc (GhcPass Parsed)))
        Actual: HsMatchContext GhcPs
    • In the first argument of ‘mkSimpleMatch’, namely ‘ctxt’
      In the expression: mkSimpleMatch ctxt pats rhs
      In an equation for ‘match’:
          match (pats, rhs) = mkSimpleMatch ctxt pats rhs
    |
669 |     match (pats, rhs) = mkSimpleMatch ctxt pats rhs
    |                                       ^^^^

src/Proto3/Suite/DotProto/Generate/Syntax.hs:673:18: error: [GHC-83865]
    • Couldn't match type: GenLocated
                             SrcSpanAnnN GHC.Types.Name.Reader.RdrName
                     with: GhcPass Parsed
      Expected: GhcPs
        Actual: HsName
    • In the ‘mc_fun’ field of a record
      In the expression:
        FunRhs
          {mc_fun = name, mc_fixity = Prefix, mc_strictness = strictness}
      In an equation for ‘ctxt’:
          ctxt
            = FunRhs
                {mc_fun = name, mc_fixity = Prefix, mc_strictness = strictness}
    |
673 |       { mc_fun = name
    |                  ^^^^

[TristanCacqueray]

This is not the prettiest thing, but at least this is now compiling with ghc-9.10.1! Marking this as ready for review.

[TristanCacqueray]

Arg, unfortunately the generated code doesn't compile, this PR produces:

@(HsProtobuf.UnpackedVec HsProtobuf.String Hs.Text) 

… instead of:

@(HsProtobuf.UnpackedVec (HsProtobuf.String Hs.Text))

Would someone know how that works?

Since that's the only failure we got, it can be fixed with: sed -e 's/HsProtobuf.String Hs.Text/(HsProtobuf.String Hs.Text)/'

[j6carey]

Probably we need to explicitly add parentheses. That isn't automatic. There are also some changes to the Nix code that will be required; I'm investigating those.

[j6carey]

@TristanCacqueray , I hope you don't mind but for reasons of speed I filed a separate PR to try to make the required Nix changes: #263 . If it works out and is approved then it should subsume this one, and I presume that as long as GHC 9.10 is correctly supported that satisfies your requirements. In any case #263 should provide some additional results from continuous integration.

[TristanCacqueray]

@j6carey thanks, feel free to assimilate this change. Note that we had to pull #252 too as it is needed for the latest nixpkgs.

[j6carey]

Note that we had to pull #252 too as it is needed for the latest nixpkgs.

@TristanCacqueray , curiously, the Nix environment I set up in #263 doesn't end up using aeson-2.2, so I didn't end up needing to incorporate those changes just to get it to build.

[j6carey]

Note that we had to pull #252 too as it is needed for the latest nixpkgs.

@TristanCacqueray , curiously, the Nix environment I set up in #263 doesn't end up using aeson-2.2, so I didn't end up needing to incorporate those changes just to get it to build.

Ah, because we did some extra work to stay on the old version. I'll think about this...

[TristanCacqueray]

Right, this aeson came with the latest nixpkgs, for the record, here is the change where we bumped to 9.10: change-metrics/monocle#1142