VSCode - Suppress Error Toast

[AThilenius]

Feature request: Add an option to suppress all Rust Analyzer errors from bubbling up to VSCode toasts.

---

Each and every single Rust Analyzer crash results in a toast being displayed in VSCode. Even if you chose to ignore the toast and leave it up, repeat toasts will play an attention-grabbing animation. I love Rust, I love Rust Analyzer, I'm immensely grateful to everyone in the world that made those two possible, but the toasts are like a splinter I just can't ignore. Maybe it's some type of OCD, but it's really painful 🙏 I brought this up in issue 717 as well.

"We should fix the bugs instead"

I get a deluge of popups when I'm trying to git rebase. As I would fully expect, git merge conflicts aren't valid Rust and I don't expect (or want) Rust Analyzer to understand them. There are an infinite number of edge cases that fit this description. Process was killed, container was closed, SSH terminated, filesystem is FUBAR... it's impossible to fix every bug. Making the server hardened against crashes is one option (IMO probably not a great option, panics are a good way to report stack traces and handle truly unrecoverable situations 'cleanly'). I don't think fixing every bug is a good substitution for hiding these toasts.

"We need bug reports"

Microsoft has a guide on user telemetry for VSCode. A few hundred people opted into telemetry is orders of magnitude more useful than hoping someone opens an issue for each of the many error toasts that popup throughout the day, many of which aren't useful because the user knows exactly why the Rust isn't valid. This kind of stuff is largely plug-and-play too, things like Sentry.io can help you filter out non-errors (like trying to parse a git merge conflict).

[Zeenobit]

I want to bump this issue because it has been slowly driving me insane lately. Usually whenever I'm developing in Bevy a chunk of my screen space is dedicated to Rust analyzer error notifications that seem to have absolutely no functional effect. I feel like lately it has just been getting worse, but I have no data to back up this claim.

Please give me my screen space and sanity back. 🥲

[Veykril]

While adding a button to the error toasts to surpress further ones for the session sounds okay, its still less than ideal that a rebase can cause panics. r-a should ideally never panic no matter what text input it gets fed.

@Zeenobit does this occur to you due to fiddling with the git repo as well, or is it just a thing with your bevy project?

[Zeenobit]

For me, it's usually just as I'm editing code. I suspect it may have something to do with lots of macro usage in Bevy, but that's just a guess. I'll try to collect some data and post any findings here.

[Veykril]

please report panics like that if you can, the point of this issue is not being able to ignore rust-analyzer from failing.

[AThilenius]

r-a should ideally never panic no matter what text input it gets fed

Not wanting a server (a persistent process) to ever panic is a good goal. I think a one-shot process panicking on garbage input is not just reasonable, it's a good design. It communicates error state clearly to a calling process. It happens in every Rust codebase I've ever worked on, often when Rust/Cargo files are in transient nonsensical states from editing.

BUT I really want to stress this again, that is NOT what this issue is about. It is about disruptive error spam in VSCode, which is against almost every single recommendation Microsoft has on VSCode notification handling.

Edit: I'll take a crack at a PR. Do you guys have any 'getting started with dev' instructions? I got the VSCode extension to build and launch but get a Bootstrap error [Error: Failed to execute rust-analyzer --version when I open a Rust file.

[flodiebold]

@AThilenius have you actually reported your crash? I know you're saying this issue isn't about that, but before you start claiming that it's impossible to address your annoyance by just fixing the crashes, you could at least try; otherwise it's just a self-fulfilling prophecy.

[Zeenobit]

@AThilenius have you actually reported your crash? I know you're saying this issue isn't about that, but before you start claiming that it's impossible to address your annoyance by just fixing the crashes, you could at least try; otherwise it's just a self-fulfilling prophecy.

I think everyone here fully understands that the correct reaction to these errors is fixing them.
This isn't about fixing rust-analyzer errors. This is about using telemetry to report these errors directly to people who are better equipped to do something about it, instead of passing it to the user and hoping they do that manually or try and fix it themselves.

Most people would simply dismiss these errors since it's very difficult (at least for me) to shift one's focus from solving a complicated programming problem to debugging rust-analyzer. So more often than not, the error message just serves to annoy the user and not provide any actual value.

[flodiebold]

Even if we introduce telemetry, it's still arguable whether we should completely hide the crash from the user. And the question is whether we really need telemetry; I would expect programmers using an open-source software to be capable of reporting crashes when they encounter them.

So more often than not, the error message just serves to annoy the user and not provide any actual value.

That's in the hands of the user who is choosing not to report the bug though. And if the IDE is just not working without an error message, can also be very annoying to users.

[Zeenobit]

I seem to be getting these 2 error notifications when moving/renaming source files and saving Cargo.toml mid-edit:

rust-analyzer failed to load workspace: Failed to read Cargo metadata from Cargo.toml file...

Followed by:

cargo check failed

Any edits which invalidate the cargo seem to cause these notifications. Why should these be reported to user as error notifications? I know the cargo is invalid, I'm in the process of fixing it.

[AThilenius]

@Zeenobit said it well.

And yes, I have. Like contributing to this issue which was opened in 2019. But once again... that is NOT what this issue is about. I'm going to keep repeating that ad nauseam 😋

you could at least try

I have a day job and very limited time to contribute to the open source community, but I contribute where I can (and am offering a PR for this with some guidance). Speaking of Bevy, I wrote the original transform hierarchy impl Bevy is based on for Legion, and I donate to the project financially. Responses like "don't be lazy and fix it yourself" aren't productive or useful, they will quickly derail this conversation. I'm not lazy, nor am I entitled about open-source. I am human and only have 24 hours in my day.

[lnicola]

I think crashes while switching git branches are an old problem and we don't gain much from getting hundreds of reports about it.

[flodiebold]

And yes, I have. Like contributing to #717 which was opened in 2019. But once again... that is NOT what this issue is about. I'm going to keep repeating that ad nauseam yum

The problem you mentioned in that issue was supposedly fixed 5 days later. Are you saying you're still encountering that crash, even with current rust-analyzer? If so, that would be important to know.

Responses like "don't be lazy and fix it yourself" aren't productive or useful, they will quickly derail this conversation. I'm not lazy, nor am I entitled about open-source. I am human and only have 24 hours in my day.

I'm sorry for not expressing myself clearly; I meant you could try reporting the issue, not fixing it. I don't expect anyone to spend time trying to fix issues they encounter, I completely understand not having the time for that; but I do expect them to at least report issues they complain about.

[flodiebold]

I seem to be getting these 2 error notifications when moving/renaming source files and saving Cargo.toml mid-edit:

rust-analyzer failed to load workspace: Failed to read Cargo metadata from Cargo.toml file...

Followed by:

cargo check failed

Any edits which invalidate the cargo seem to cause these notifications. Why should these be reported to user as error notifications? I know the cargo is invalid, I'm in the process of fixing it.

Completely agree, these should be displayed as some kind of persistent error state instead of a notification.

[AThilenius]

I'm offering a PR to disable repeat errors if someone will help me get a development environment setup/offer guidance, and if the PR will be accepted. I am not offering to be a human telemetry gathering tool. Add telemetry if you need telemetry.