LSP errors disappear after file save and don't come back for a few seconds

[dam5h]

Summary

I often use spc-D to view all errors or clippy warnings for my rust workspace from helix.

I use this to walk around the files and fix them one at a time. Recently I noticed that as soon as I fix one error and then hit spc-D again all the other errors have vanished and I just get a blank list of errors. If I then wait a couple seconds and re-query, I get them all again.

This used to be essentially instant, and I am not sure what changed. I did upgrade Helix a few times and I honestly am not sure when the behavior started getting worse unfortunately.

Is anyone else seeing this?

Reproduction Steps

Here I see my list of all issues in the workspace:

Then I edit a single one to fix it and then I requery for the same list:

I would expect it to be nearly identical with only one change. I noticed just adding a line to the file retriggers the whole thing and I get a partial list. Maybe this was always firing off a full on re-incremental compilation from rust-analyzer but something seems much slower now.

Sometimes it just never actually shows up but if I wait a while and then close the query and re-type spc-D they all show up, so it needed the quit and re-query to render.

Helix log

~/.cache/helix/helix.log

2024-05-02T22:40:12.653 helix_term::application [DEBUG] received editor event: Redraw
2024-05-02T22:40:12.655 helix_view::document [DEBUG] id 1 modified - last saved: 0, current: 0
2024-05-02T22:40:12.655 helix_view::document [DEBUG] id 2 modified - last saved: 1, current: 1
2024-05-02T22:40:14.108 helix_lsp::transport [INFO] rust-analyzer <- {"jsonrpc":"2.0","method":"$/progress","params":{"token":"rust-analyzer/flycheck/0","value":{"kind":"end"}}}
2024-05-02T22:40:14.108 helix_term::application [DEBUG] received editor event: LanguageServerMessage((LanguageServerId(1v1), Notification(Notification { jsonrpc: Some(V2), method: "$/progress", params: Map({"token": String("rust-analyzer/flycheck/0"), "value": Object {"kind": String("end")}}) })))
2024-05-02T22:40:14.142 helix_term::application [DEBUG] received editor event: Redraw
2024-05-02T22:40:14.143 helix_view::document [DEBUG] id 1 modified - last saved: 0, current: 0
2024-05-02T22:40:14.144 helix_view::document [DEBUG] id 2 modified - last saved: 1, current: 1
2024-05-02T22:40:14.515 helix_view::document [DEBUG] id 1 modified - last saved: 0, current: 0
2024-05-02T22:40:14.516 helix_view::document [DEBUG] id 2 modified - last saved: 1, current: 1
2024-05-02T22:40:14.766 helix_term::application [DEBUG] received editor event: IdleTimer
2024-05-02T22:40:15.176 helix_view::document [DEBUG] id 1 modified - last saved: 0, current: 0
2024-05-02T22:40:15.177 helix_view::document [DEBUG] id 2 modified - last saved: 1, current: 1
2024-05-02T22:40:15.426 helix_term::application [DEBUG] received editor event: IdleTimer
2024-05-02T22:40:15.645 helix_view::document [DEBUG] id 1 modified - last saved: 0, current: 0
2024-05-02T22:40:15.646 helix_view::document [DEBUG] id 2 modified - last saved: 1, current: 1
2024-05-02T22:40:15.806 helix_view::document [DEBUG] id 1 modified - last saved: 0, current: 0
2024-05-02T22:40:15.806 helix_view::document [DEBUG] id 2 modified - last saved: 1, current: 1
2024-05-02T22:40:16.709 helix_view::document [DEBUG] id 1 modified - last saved: 0, current: 0
2024-05-02T22:40:16.709 helix_view::document [DEBUG] id 2 modified - last saved: 1, current: 1

Platform

Linux

Terminal Emulator

Kitty

Installation Method

source, main branch

Helix Version

helix 24.3 (cfca308)

[pascalkuthe]

Diagnostics are send by the lsp so this would be ra sending us ab empty diagnostic list and then a non-empty one a while after.

This is probably a case of a timeout in RA that occurs if cargo takes too long. You may want to look into build times/if something with your setup isn't correct or report a bug to RA upstream (but make sure you have a proper reproduction case and tested with the latest nightly).

Space-d nit refreshing is as intended. The picker represents a snapshot, it's not feasible to live respond if the LS updates the diagnostics in the background. I don't think that's really useful/necessary anyway since usually lsps don't send diagnostics after long delays like this

[dam5h]

Sounds good. Thank you @pascalkuthe , will look into RA side of this. Cheers!