AbortAddon: setTimeout doesn't work in combination with own AbortController instance

[adriaanmeuris]

Timeouts and aborting requests work fine when using .controller:

// Aborts after 1 second timeout
const [controller, w] = wretch(url).addon(AbortAddon()).get().setTimeout(1000).controller();

// Aborts after 0.5 second using `.abort`
const [controller, w] = wretch(url).addon(AbortAddon()).get().setTimeout(1000).controller();  
setTimeout(() => { controller.abort(); }, 500);

However when you want to provide a controller yourself, the implementation is broken:

// Aborts after 1 second timeout, but can't be aborted before because we don't provide a controller
await wretch(url).addon(AbortAddon()).get().setTimeout(1000);

// Doesn't abort after 0.5 second timeout (even though configured so), but can be aborted using `.abort`
const controller = new AbortController();
await wretch(url).addon(AbortAddon()).signal(controller).get().setTimeout(500);
setTimeout(() => { controller.abort(); }, 1000);

Additionally, I don't this it's currently possible to differentiate between a configured timeout error, and manually aborting it (they both trigger the onAborted catcher)

[elbywan]

Hey @adriaanmeuris,

// Doesn't abort after 0.5 second timeout (even though configured so)
await wretch(url).addon(AbortAddon()).signal(controller).get().setTimeout(500);

I believe you have to explicitly pass your custom controller as an argument: .setTimeout(500, controller)

If you use a custom AbortController associated with the request, pass it as the second argument.

https://elbywan.github.io/wretch/api/interfaces/addons_abort.AbortResolver.html#setTimeout

I don't this it's currently possible to differentiate between a configured timeout error, and manually aborting it

They are both equivalent indeed, since setTimeout calls controller.abort() under the hood.

[adriaanmeuris]

Thanks for clarifying. Is .signal(controller) still required if the controller is passed as a second argument to setTimeout?

I need to distinguish manual aborts (e.g. aborting irrelevant requests before they resolve) from actual timeouts, especially for error reporting. Any advice on this, or would you consider a PR that allow to explicitly separate the two cases?