netutil: consistently format ipv6 addresses

[Kidsan]

This formats ipv6 addresses to ensure they can be compared safely

Fixes #15127

[ahrtr]

LGTM

Thank you @Kidsan

[ahrtr]

Could you backport this PR to 3.4 and 3.5 as well?

[serathius]

Should we add a e2e test for IPv6?

[ahrtr]

Should we add a e2e test for IPv6?

Adding e2e test is good, but I think it can be a separate PR.

[serathius]

Makes sense, however let's posmone backporting until we ensure that overall issue of IPv6 is fixed and tested. Please remember that we are backporting to fixes to issues, not just backporting PRs for sake of backporting.

[ahrtr]

Makes sense, however let's posmone backporting until we ensure that overall issue of IPv6 is fixed and tested. Please remember that we are backporting to fixes to issues, not just backporting PRs for sake of backporting.

Do not understand the concern.

1. Do not literally compare IPv6 addresses #15127 is a real issue to me, please let me know if you don't think it's an issue. The contributor fixes the issue and backport it to 3.5 and 3.4. It makes perfect sense to me. Thx @Kidsan
2. We can add whatever e2e test for IPv6 in separate PRs. If we see any other issues, we can just continue fix them. etcd shouldn't care about the IP type (IPv4 or IPv6), so I don't worry about any big changes due to any new issues caused by IPv6.

[serathius]

For me a feature like IPv6 support cannot exist without tests. If IPv6 formatting didn't work, means that our tests were insufficient, ergo there might be other things that are in out blind spot.

Discovering and fixing one detail after another is insufficient to make sufficient progress. We need more holistic approach. My thinking is that if IPv6 had such a rudimental bug like this one, it was never properly tested. We should revisit this feature and add sufficient testing to bring up the quality of this area.

As agreed in https://github.com/etcd-io/etcd/blob/main/Documentation/contributor-guide/features.md all stable features should have a proper testing. If we don't think that IPv6 support is mature enough we should mark it as experimental.

[ahrtr]

I do not see any strong reason to reject this minor bug fix.

[sanmai-NL]

I think you can both merge this PR and open/request an issue or PR for the tests.

[serathius]

FYI I'm not blocking merging of this PR, just backporting before whole issue is resolved.