Fix iserv-proxy build on windows

[alexfmpe]

Not sure why this wasn't a problem before.
Maybe because haskell.nix does per-component builds so building the host side for windows basically never happened?

Fixes

Building executable 'iserv-proxy' for iserv-proxy-9.3...
[1 of 1] Compiling Main             ( Main.hs, dist/build/iserv-proxy/iserv-proxy-tmp/Main.o )
Main.hs:94:24: error: [GHC-39999]
    • No instance for ‘Read
                         (ghc-internal-9.1202.0:GHC.Internal.Ptr.Ptr ())’
        arising from a use of ‘read’
    • In the expression: read arg0
      In an equation for ‘wfd1’: wfd1 = read arg0
      In a stmt of a 'do' block:
        let wfd1 = read arg0
            rfd2 = read arg1
            ip = arg2
            port = read arg3
   |
94 |             let wfd1 = read arg0
   |                        ^^^^

Main.hs:109:8: error: [GHC-39999]
    • No instance for ‘Integral
                         (ghc-internal-9.1202.0:GHC.Internal.Ptr.Ptr ())’
        arising from a use of ‘fromIntegral’
    • In the second argument of ‘printf’, namely
        ‘(fromIntegral rfd2 :: Int)’
      In the second argument of ‘($)’, namely
        ‘printf
           "GHC iserv starting (in: %d; out: %d)\n" (fromIntegral rfd2 :: Int)
           (fromIntegral wfd1 :: Int)’
      In a stmt of a 'do' block:
        when verbose
          $ printf
              "GHC iserv starting (in: %d; out: %d)\n" (fromIntegral rfd2 :: Int)
              (fromIntegral wfd1 :: Int)
    |
109 |       (fromIntegral rfd2 :: Int) (fromIntegral wfd1 :: Int)

[alexfmpe]

getGhcHandle on windows is HANDLE -> IO Handle where HANDLE = Ptr () rather than CInt -> IO Handle on unix

readGhcHandle handles both cases

[alexfmpe]

Old: GHC iserv starting (in: 79; out: 78)
New: GHC iserv starting (in: {handle: <file descriptor: 79>}; out: {handle: <file descriptor: 78>})

Is preserving the old behavior on unix worth CPP?

[angerman]

@alexfmpe you mean the output only? I don't think it is. Iserv-proxy and interpreter are primarily for functionality. I hope that with the cross support in ghc we can setup proper testing here.

[angerman]

@hasufell do you have any thoughts on this change?

@hamishmack do you know why we don't run into this with Haskell.nix?