Error not reported/recognised

[ivan-m]

I tried to load a file (using cabal repl) and the message buffer just said it was compiling a file without changing (when it's normally instant). The interactive buffer never said anything.

When I looked at the process log, this was the contents:

/home/ivan/Haskell/paraphrase/src/Text/Paraphrase/Types.hs:(336,47)-(337,78):
    Couldn't match expected type ‘ParseError s -> Result s r’
                with actual type ‘Result s r’
    Relevant bindings include
      e :: ParseLog s
        (bound at /home/ivan/Haskell/paraphrase/src/Text/Paraphrase/Types.hs:334:31)
      pSt' :: ParseState s
        (bound at /home/ivan/Haskell/paraphrase/src/Text/Paraphrase/Types.hs:334:26)
      fl' :: ParseState s -> ParseLog s -> ParseError s -> Result s r
        (bound at /home/ivan/Haskell/paraphrase/src/Text/Paraphrase/Types.hs:334:22)
      sc' :: ParseState s -> a -> Result s r
        (bound at /home/ivan/Haskell/paraphrase/src/Text/Paraphrase/Types.hs:343:22)
      restoreAdd :: ParseState s -> ParseState s
        (bound at /home/ivan/Haskell/paraphrase/src/Text/Paraphrase/Types.hs:346:22)
      sc :: Success s a r
        (bound at /home/ivan/Haskell/paraphrase/src/Text/Paraphrase/Types.hs:333:29)
      (Some bindings suppressed; use -fmax-relevant-binds=N or -fno-max-relevant-binds)
    In the expression:
      mergeIncremental pSt pSt' $ \\ pSt'' -> runP p2 pSt'' fl sc
    In an equation for ‘fl'’:
        fl' pSt' e
          | isCommitted pSt' = failure (restoreAdd pSt') e
          | otherwise
          = mergeIncremental pSt pSt' $ \\ pSt'' -> runP p2 pSt'' fl sc
    In the expression:
      let
        fl' pSt' e
          | isCommitted pSt' = failure (restoreAdd pSt') e
          | otherwise = mergeIncremental pSt pSt' $ \\ pSt'' -> ...
        sc' pSt' = sc (restoreAdd pSt')
        restoreAdd pSt' = pSt' {add = add pSt <> add pSt'}
      in ignoreAdditional pSt $ \\ pSt' -> runP p1 pSt' fl' sc'

It seems for some reason this error couldn't be parsed/recognised; I don't know if it's the error type or its length.

[bmmoore]

The problem is the error range format "(336,47)-(337,78)". Haskell-mode so far only recognizes lines:column and line:column-column.
Can you get GHC to produce an error message using that format when compiling the file outside of haskell-mode?
I'm working a bit to fix this, but so far I just have a regex that makes those errors be displayed properly. Other functionality is still broken, like actually jumping to them, or removing them from the repl when they are resolved.

[chrisdone]

Yes, someone please provide a reproducible self-contained test case for this. Updating the parsing is straight-forward, but I need a real example.

[bmmoore]

I've got reproducible, still working on the self-contained part. I have a file that produces messages of that form if I drop it into a large sandboxed cabal project I'm working on. No success so far setting up smaller project where that file shows any but one-line error locations.

[bmmoore]

Now that I finally thought to check ps, I see the process is running with a -ferror-spans option. Running ghci -ferror-spans on 7.8.2 even simple things like this will produce such an error:

class Test a b where
  method :: a -> b

As the entire contents of a test.hs, invoking ghci -ferror-spans test.hs produces

GHCi, version 7.8.2: http://www.haskell.org/ghc/  :? for help
Loading package ghc-prim ... linking ... done.
Loading package integer-gmp ... linking ... done.
Loading package base ... linking ... done.
[1 of 1] Compiling Main             ( test.hs, interpreted )

test.hs:(1,1)-(2,18):
    Too many parameters for class ‘Test’
    (Use MultiParamTypeClasses to allow multi-parameter classes)
    In the class declaration for ‘Test’
Failed, modules loaded: none.

[chrisdone]

I think this is fixed in c72cc06

[gracjan]

@ivan-m, @bmmoore, @chrisdone: is this one working for you? Can be closed?

[ivan-m]

Yeah, I think so.