Thank you for the bug report!

I can not reproduce on arch linux with HLS 2.9.0.1 installed via ghcup.
However, I can reproduce the issue with a nixpkgs installed HLS binary.

I am able to reproduce the issue using a nixpkgs installed HLS binary as well as on arch linux with HLS 2.9.0.1 ghc 9.6.6.

I have the same issue in doomemacs, emacs+lsp-mode , vanilla arch, ghc 9.10.1

Same here, vanilla emacs+lsp-haskell, macOS Ventura 13.7.1, ghc 9.4.8, hls 2.9.0.1

I’m seeing the same issue in vanilla Emacs with lsp-haskell, Arch Linux (well, EndeavourOS), GHC 9.4.8, HLS 2.1.0.1

Reading into the code, I don't think there is anything wrong, sometimes additional information can be requested, and sometimes the server doesn't have any additional information. E.g., documentation or type signatures for non-local identifiers.
We can't be that fine-grained in the Protocol, so the client always tries to resolve completion items which sometimes "fails" but the server can still use what they already have.

Is the current behaviour annoying, e.g., does Emacs or vim spam you with errors?

yeah in emacs, its extremely annoying. a buffer with that error keeps popping up as you type

But this must be a new change in emacs, right? I think we had this behaviour for a while and just now we started to get these reports?

Seemingly, HLS is behaving correctly. I think this issue is resolvable by extending CompletionResolveData with a NothingToResolve and inserting as the default for all completion items.

Is the current behaviour annoying, e.g., does Emacs or vim spam you with errors?

Correct. I get a message every time the autocomplete event fires, i.e. every time I press a key (or nearly so).

message every time the autocomplete event fires, i.e. every time I press a key (or nearly so).
That's exactly what's happening. Extremely annoying. I limited minibuffer height with 1 line, but in this case i don't see meaningful messages / menus from other plugins.

I created #4463 which I will sneak into #4448 to get it out ASAP.

Does someone want to double check this works?

Does someone want to double check this works?

If you tell me what to do - i'll try :-)

If you use GHCup, the following works:

ghcup compile hls -g d51c68dcf7a418bde4d02f62345710f298168328 --ghc 9.6.6

Or whatever GHC version you use in your project.

You can also use clone the project, and run cabal install, then point emacs to your freshly installed HLS version.

If you use GHCup, the following works:

ghcup compile hls -g d51c68dcf7a418bde4d02f62345710f298168328 --ghc 9.6.6

I did this for ghc 9.10.1
This message now doesn't appear when auto-completion list consist of haskell only keywords / modules etc, but if it includes one of my function in it - this message still continue to appear. Screenshots below

Thanks, another test to add to the testsuite!

@dmitrykvasnikov Where was the local function defined you tried to complete? In the same module, different module, etc...? I am unable to reproduce the issue.

@fendor
In this same module (it's written above on screenshot). Just tried with function defined in different module - works fine, no message.
Just now I did as example:
func :: Int -> Int -> Int
func = (+)
So if after that i put in my code:
g = fu .. - in this moment autocomplete gives me suggestions (with function defined above) and error message appears.

Can't reproduce, I am getting the correct message

Trace - 17:31:55] Received response 'completionItem/resolve - (20)' in 2ms.
Result: {
    "data": {
        "itemFile": "file:///home/hugin/Documents/haskell/hls-completion-resolve/A.hs",
        "itemName": [
            "A",
            "main-398bd560290e6bafe7a4d6959f05d45a6937707a",
            "v",
            "func"
        ],
        "itemNeedsType": false,
        "tag": "CompletionResolveData"
    },
    "detail": ":: Int -> Int -> Int",
    "documentation": {
        "kind": "markdown",
        "value": "*Defined at line 3, column 1 in this module*\n"
    },
    "insertText": "func",
    "insertTextFormat": 2,
    "kind": 3,
    "label": "func",
    "sortText": "00"
}

Maybe the compiled binary wasn't used? I am double checking my instructions.

I am a little bit confused 😅 After running ghcup compile ..., does ls -alh ~/.ghcup/bin/haskell-language-server-9.10.1~2.9.0.1 hint that the binary was freshly created?
Does ~/.ghcup/bin/haskell-language-server-9.10.1 point to ~/.ghcup/bin/haskell-language-server-9.10.1~2.9.0.1?

I'm compiling this myself to see if I can reproduce (taking a while)

ok, so if I'm doing this correctly. the error still pops up, but less frequently. not so helpful footage of when it triggers attached

using ghc 9.6.6

checked the binary is freshly created

>ls -alh ~/.local/bin/haskell-language-server
Permissions Size User  Date Modified Name
lrwxrwxrwx     - bwkam 10 Dec 20:08   /home/bwkam/.local/bin/haskell-language-server -> ../state/cabal/store/ghc-9.6.5/haskell-language-server-2.9.0.1-e-haskell-language-server-bda016429039356b5e7ae15d0b54c3a626539f8f90dccf348e7458730e7bef3d/bin/haskell-language-server

logs:

2024-12-10T18:29:54.785159Z | Warning | No plugin handles this "completionItem/resolve" request.
2024-12-10T18:29:54.786113Z | Warning | No plugin handles this "completionItem/resolve" request.
2024-12-10T18:29:55.405965Z | Warning | No plugin handles this "completionItem/resolve" request.
2024-12-10T18:29:58.902473Z | Info | Live bytes: 20.24MB Heap size: 662.70MB
2024-12-10T18:30:03.242066Z | Info | Could not identify reverse dependencies for NormalizedFilePath "/home/bwkam/Documents/dev/aoc24/app/Main.hs"
2024-12-10T18:30:06.793302Z | Warning | No plugin handles this "completionItem/resolve" request.
2024-12-10T18:30:06.793467Z | Warning | No plugin handles this "completionItem/resolve" request.
2024-12-10T18:30:08.573218Z | Warning | No plugin handles this "completionItem/resolve" request.
2024-12-10T18:30:08.573398Z | Warning | No plugin handles this "completionItem/resolve" request.
2024-12-10T18:30:09.078120Z | Warning | No plugin handles this "completionItem/resolve" request.

I am a little bit confused 😅 After running ghcup compile ..., does `` hint that the binary was freshly created? Does ~/.ghcup/bin/haskell-language-server-9.10.1 point to `/.ghcup/bin/haskell-language-server-9.10.12.9.0.1`?

I've just checked - yes to both questions. haskell-language-server-wrapper updated as well (if it's relevant) .

Very puzzling... the Completions shown by @bwkam look completely wrong, too. We should never suggest LANGUAGE pragmas in this context as a possible completion?!

yeah, I was not sure why it was suggesting that