bug: aborted signal not handled properly by libp2p router

[SgtPooki]

See https://github.com/ipfs/service-worker-gateway/pull/681#issuecomment-2816001952.}

I don't have a good repro usecase besides some code changes in ipfs/service-worker-gateway#681 and manual testing.

It's difficult to set up tests with all the things enabled that I need..

It's somewhere in the combination of ipfs-unixfs-exporter and @helia/routers when bitswap is enabled.

I was able to get this fixed in verified-fetch with:

import { abortableSource } from 'abortable-iterator'

// ...

	const iterator = options?.signal != null 
		? abortableSource(exporterWalk(path, blockstore, options), options.signal) 
		: exporterWalk(path, blockstore, options)
    for await (const entry of iterator) {

// ...

[SgtPooki]

I put together https://github.com/SgtPooki/repro-helia-libp2p-router-abort-handling and am trying to reproduce this error but have not been having great luck :/

[SgtPooki]

I need to revert ipfs/helia-verified-fetch#214 once we figure out what is actually happening

[2color]

I also just noticed this today when publishing an IPNS record with @helia/ipns

https://github.com/ipfs-examples/helia-examples/blob/f9ed78fa2b7fd9e0ee9f2fdc1b99eb597e5cdc00/examples/helia-101/501-ipns.js#L30-L40

If you set the abort signal to 1 second, the promise doesn't get rejected.

I think that would confirm that this might be related to the router (as you suspected)

[2color]

One source of this bug has been fixed in a dependency of js-libp2p (see achingbrain/it#159 and libp2p/js-libp2p#3094)

However, the bug persists sometimes in dht.provide operations.

As mentioned above, it can be easily reproduced in the following examples:

• https://github.com/ipfs-examples/helia-examples/blob/f9ed78fa2b7fd9e0ee9f2fdc1b99eb597e5cdc00/examples/helia-101/501-ipns.js#L30-L40
• https://github.com/ipfs-examples/helia-examples/pull/465/files#diff-3cf9da14179b49a9a3b934734051568058cdf5ba77b2b3052e4f7b29c7ce5487R34-R36