Fork bomb with rust-analyzer clippy

[RoloEdits]

Summary

I had noticed this once before but though it was something with my own setup, but saw this reddit post and thought I would make an issue of it, incase its something more.

After opening up the zed repo and trying to check the source code, for example the breadcrumbs.rs, and doing a hover action, it eventually begins to lag as clippy processes are started over and over.

Reproduction Steps

Helix log

https://gist.github.com/RoloEdits/cbabfa4ab230019d8795111c69029063

Platform

Windows

Terminal Emulator

wezterm 20241205-214838-42ecaa64

Installation Method

source

Helix Version

helix 25.1 (2bdb710)

[archseer]

Is this a rust-analyzer issue?

[RoloEdits]

I was thinking that as well, but wanted to try to get one place that isnt reddit to discuss, as one person said they were in a vue project and this happened. But seems no one came here to share details in the meantime. I'll open up an issue with RA.

[ngn999]

After open zed master in helix( 25.01 (e698b20)), cpu 100%:

It looks like another issue.

[RoloEdits]

The OP of the reddit post wasnt using rust-analyzer. So it could speak to broader issues.

Did you open up any rust codebase prior to that?
-me

I am a C++ developer :p
- OP

[pascalkuthe]

Are you using autosave? I would guess it's multiple save requeat running in parallel/not keeping up. In general we dont control when rsut-analayzer runs clippy (it runs on save by default) so I can only imagine us triggering more save notifications

[RoloEdits]

Im not explicitly using it in my config. As far as timeouts are concerned I have this.

idle-timeout = 50
completion-timeout = 5

[pascalkuthe]

We don't use the idle timeout as much anymore but does it also happen without those low timeouts?

[RoloEdits]

Commenting it out (so default) it still happens.

[pascalkuthe]

I suspect this is due to the fact that we register a new root folder for every file we opent gat is in a root we dint have available yet (from looking at the source).

In first startup it sgould only be running the check once but if you work for a while and open more dependencies I guess you accumulate workspaces which would each run a check command.

This didn't happen in the past but I suspect that RA has changed behavior. I recall it being mentioned in the changeling. I think the behavior was reverted recently so I winder if it's still there.

Workspace root detection is a huge pain since lsp is based around the fact that the workspace root is simply the directory where you open vscode.

In general it's not easily possible to distinguish dependencies from additional workspaces. I have a heuristic in kind but before doing any work would like to confirm first that this is what causes the issue (and that it's not tied with latest RA nightly).

[RoloEdits]

The tracking issue for RA: rust-lang/rust-analyzer#18851