We need to add configuration option (Flag) Swarm.RelayService.FooLimitEnabled that is

• true by default (implicit limits), but
• when explicitly set to false ensures we initialize RelayService with WithInfiniteLimits from this PR

This is unrelated to this PR, if I understand correctly. Kubo didn't have an option to enable a v1 relay - that's why we introduced the relay daemon.

@marten-seemann ack, I'll add config for libp2p/go-libp2p#2125 in a separate PR, after this one is merged.

@marten-seemann can you rebase on top of the latest master? I'll take a look in a bit. (I can also do it).

Updated to the official v0.26.0 release. This should be ready to go now :)

Things still to fix (I will take a look, but may take longer time as I'm not very knowledgeable about this codebase):

• Remove Circuit V1 Relay tests from interop as it was removed.
  • fix: remove relay v1 go tests interop#592 needs to be merged and released.
  • Version should be updated here.
• Update sharness tests with the new metrics included by go-libp2p v0.26.0
• kad-dht tests still failing (hoping that will be fixed with chore: bump go-libp2p-kad-dht to v0.21.1 #9663)

• Update sharness tests with the new metrics included by go-libp2p v0.26.0

Not my business, but I've never understood why Kubo would make any assertions on the metrics exposed by libp2p. Maybe consider filtering out all libp2p_ metrics?

@marten-seemann yeah, definitely. I was not insinuating you should make the changes. I will take a look at that. Filtering them out seems like a better idea.

Edit: I will keep them as of now as there's some other metrics that are not entirely dependent on Kubo, but on other imported packages. For consistency, it's better to keep them all. Let's just hope I added all the remaining metrics on cb5563f. Turns out these sharness tests are not macOS friendly and it's a hassle to test them.