[cruspy2004]

Closes #4181

ImpitHttpClient hands impit only the caller's abort signal, so impit's own 30 s default timeout still applies underneath it. With navigationTimeoutSecs: 60, a page that answers after 35 s fails at 30 s with "Request timed out after 60 seconds.", and extendTimeout() can't get past it either.

When an abort signal is passed, the client now gives impit a timeout of about 24 days, so the signal alone decides when the request ends. impit has no way to disable its timeout (0 and Infinity both time out immediately), hence the large value. Without a signal, impit's default is unchanged.

Against a local server that answers after 35 s, the page is now handled at ~36 s with navigationTimeoutSecs: 60, and still fails at 5 s with "Navigation timed out after 5 seconds." when set to 5.

[nikitachapovskii-dev]

Thanks! Tested locally and the issue sample works now

A few things though:
FileDownload still gets cut at 30 s. It doesn’t pass a signal, so impit keeps its default.
If the navigation window runs out while the body is still downloading, the download now just keeps going. Before, impit’s 30 s stopped it.
sendRequest with a signal but no deadline can now hang forever. Before, it failed at 30 s.

[cruspy2004]

Thanks for testing!

• Download after the window runs out: right. On abort impit stops the body read but keeps the connection downloading. 7e8ad92 pipes the body through the signal, so an abort now cancels the download as well. Against a slow-streaming server, the connection now closes right after the abort.
• FileDownload: its request runs in the context pipeline outside any timeout frame, so there is no task signal to pass yet. Giving downloads a deadline means choosing one, which felt out of scope here, but I'm happy to follow up if you have a preference.
• A signal without a deadline: that matches FetchHttpClient, since fetch has no default timeout either. Callers who want a limit can pass timeoutMillis, which gets combined into the signal.

[barjin]

lgtm, thank you @cruspy2004! I pushed two commits on top: the switch to impit's new timeout: null, and a fix for HttpCrawler, where the body download kept going after a failed request (the request's signal never aborted once the headers were in). Hope you don't mind 🙏 FileDownload as a follow-up is fine by me.