Trying out haskell-lsp 0.14

[alanz]

Tests failing, suspect a JSON encoding problem.

ping @cocreature, see the circle logs

Locally I get the first two of many failing tests, all with a spurious <0>

      test/functional/FunctionalCodeActionsSpec.hs:557:9: 
      1) FunctionalCodeActions, code actions, import suggestions, formats with brittany, Execute HsImport with formatter brittany, works with 3.8 code action kinds
           expected: "Import module Control.Monad (when)"
            but got: "Import module Control.Monad (<0>when)"
    
      To rerun use: --match "/FunctionalCodeActions/code actions/import suggestions/formats with brittany/Execute HsImport with formatter brittany/works with 3.8 code action kinds/"
    
      test/functional/FunctionalCodeActionsSpec.hs:598:16: 
      2) FunctionalCodeActions, code actions, import suggestions, formats with brittany, Execute HsImport with formatter brittany, import-list formats
           Actual list is not a permutation of expected list!
             expected list contains:   ["import qualified Data.Maybe", "import           Control.Monad                  ( when )", "main :: IO ()", "main = when True $ putStrLn \"hello\""]
             actual list contains:     ["import qualified Data.Maybe", "import Control.Monad ((<0>when))", "main :: IO ()", "main = when True $ putStrLn \"hello\""]
             the missing elements are: ["import           Control.Monad                  ( when )"]
             the extra elements are:   ["import Control.Monad ((<0>when))"]
    
      To rerun use: --match "/FunctionalCodeActions/code actions/import suggestions/formats with brittany/Execute HsImport with formatter brittany/import-list formats/"

[fendor]

Did these errors happen before #1284 had been merged?

[alanz]

@fendor good point, I had forgotten about that. We will have to see what circle says about the merge result. I assume it was fine, as it showed green on CI before I merged, and you had updated it recently.

[cocreature]

Hm, it looks unrelated to the changes I made but I could very well be wrong.

[alanz]

It looks like circle was happy with @fendor 's merge: https://circleci.com/workflow-run/fc727e9b-c6a9-4ed6-9f27-9f6332e248e8
I am building it locally too, to check, then will investigate the rest.

There is also an updated lsp-test, perhaps @bubba knows something about that.

[alanz]

Ok, seems there is something whacky on my local machine.

[lukel97]

The functional tests for HIE passed for me on my own branch with LSP 0.14, but could be an OS specific thing

…

On 13 Jun 2019, at 15:35, Alan Zimmerman ***@***.***> wrote: It looks like circle was happy with @fendor 's merge: https://circleci.com/workflow-run/fc727e9b-c6a9-4ed6-9f27-9f6332e248e8 I am building it locally too, to check, then will investigate the rest. There is also an updated lsp-test, perhaps @bubba knows something about that. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub, or mute the thread.

[alanz]

I am on linux too. But been messing with cabal builds, and ghc 8.8., etc. Could have a whacky config by now.

So I reckon just merge and see what happens. Circle is happy.

[alanz]

I did a bisect on this, and the problem comes from 3350686

$ git bisect log
git bisect start
# good: [83c28a907eda2a68aeeb254f8baa9f3add08aacf] Merge pull request #1287 from alanz/hlint-2.1.24
git bisect good 83c28a907eda2a68aeeb254f8baa9f3add08aacf
# bad: [d679eb595a6a1583a8b63a3fdd5964f84a8d00f6] Merge pull request #1288 from alanz/haskell-lsp-0.14
git bisect bad d679eb595a6a1583a8b63a3fdd5964f84a8d00f6
# bad: [988dcff2eb18edcee0bbed5a10e597f9ca9d1ad9] Add space to code action title
git bisect bad 988dcff2eb18edcee0bbed5a10e597f9ca9d1ad9
# bad: [3835a09db92c3a3c4f0761e50cd22e70f191c603] Fix tests, add documentation and add newtypes
git bisect bad 3835a09db92c3a3c4f0761e50cd22e70f191c603
# bad: [07822b472e0495a1819008362e53ad07e670fdc6] Add documentation for hsimport plugin
git bisect bad 07822b472e0495a1819008362e53ad07e670fdc6
# bad: [3350686dc36c9ed012a817b6806ec1a34c4b2d05] Clone HsImport API for more fine grained control
git bisect bad 3350686dc36c9ed012a817b6806ec1a34c4b2d05
# first bad commit: [3350686dc36c9ed012a817b6806ec1a34c4b2d05] Clone HsImport API for more fine grained control

I think that what is happening is that I have a full hoogle db locally, we don't have it on the CI box.

And we need to add something like https://github.com/haskell/haskell-ide-engine/blob/master/src/Haskell/Ide/Engine/Plugin/Hoogle.hs#L137 into it somewhere:

        unHTML = T.replace "<0>" "" . innerText . parseTags

cc @fendor

[alanz]

And after deleting and re-creating my ~/.hoogle directory as per the circleci config, the tests pass for me.

I tried creating additional hoogle generate dbs, but it still passes. I think there is a landmine waiting out there.

[fendor]

I have a good idea where this might be coming from. Hoogle searchPackage' defines almost exactly this function. In HsImport importListActions also obtains the function name by splitting "::". That may be wrong.

[fendor]

I think, the fix would be to adapt searchPackage' to use unHtml that will be merged via #1290