Use Tokio::spawn instead of getting an executor handle (#36701)

This was causing panics due to the handles being dropped out of order.
It doesn't seem possible to guarantee the correct drop ordering given
that we're holding them over await points, so lets just spawn on the
tokio executor itself which gives us access to the state we needed those
handles for in the first place.

Fixes: ZED-1R

Release Notes:

- N/A

Co-authored-by: Conrad Irwin <conrad.irwin@gmail.com>
Co-authored-by: Marshall Bowers <git@maxdeviant.com>
This commit is contained in:
Julia Ryan 2025-08-21 12:19:57 -05:00 committed by GitHub
parent d166ab95a1
commit 1b2ceae7ef
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
5 changed files with 38 additions and 18 deletions

View file

@ -102,13 +102,7 @@ impl CloudApiClient {
let credentials = credentials.as_ref().context("no credentials provided")?;
let authorization_header = format!("{} {}", credentials.user_id, credentials.access_token);
Ok(cx.spawn(async move |cx| {
let handle = cx
.update(|cx| Tokio::handle(cx))
.ok()
.context("failed to get Tokio handle")?;
let _guard = handle.enter();
Ok(Tokio::spawn_result(cx, async move {
let ws = WebSocket::connect(connect_url)
.with_request(
request::Builder::new()