> I think that was because IPv6 support had been incorrectly configured
> for Docker, perhaps that is the same thing happening in your setup?

In the same docker, the PR service works with PRSERV_HOST =
"localhost:0" before, but it begins not work recently after some code
restructure for prserv part.

BTW, the ipv6 is disabled in the docker of my env.

Check your /etc/hosts

There is ::1 localhost localhost-ip6 even when ipv6 is disabled in docker's deamon.json.

And that makes bitbake (asyncio) to try binding to ::1 instead of 127.0.0.1, if you remove "localhost" from ::1 in /etc/hosts in running container, then it works again OK.

My patch allows you to use 127.0.0.1:0 explicitly and still keep the same behavior (as autostarted prserv with random port) as when localhost:0 is being used (and also for consistency).

Your patch is IMHO good addition to that (that the localhost is resolved in python and then people don't need to use 127.0.0.1 explicitly).

Regards,