I'm with Timo and Kunal here, we should've disabled the dependent pipeline ages ago. It doesn't scale and the cases it catches are minimal in practice versus the disruption it causes.

Does it work for the old site that was moved now? If so I'd say retitle and call it fixed.

This is mostly documented on wikitech, but I'll provide a few more in depth pointers here. Luckily there's no schema changes on this release so you can skip the init steps.

Yes

Have you looked into protocol buffers?

Teaching to not use -A is good. Setting it in your ~/.ssh/config is useful in case it's not disabled system-wide by default :)

As I've said before, there's no need to push them?

In T198138#4313147, @Krenair wrote:

Might be better to ensure that all privileged users at least know that they should never use -A to a host they do not fully trust, and that labs machines must be treated as untrustworthy.

I already granted him access some time ago to chairwiki.

I re-ran it. It dies with an exception anyway :P

I could've sworn we ran this after we changed that setting...

How about https://gerrit-review.googlesource.com/c/gerrit/+/170693?

Done & announced

Removing from 1.31 blockers.

Was this a problem for the rc's?

We haven't needed to replicate any refs other than heads and tags since we brought gitiles online... Disable them. Now. And prune them from Phab while we're at it.

You can't have an interactive shell. But sync-with-gerrit doesn't do that....try appending gerrit ls-projects to that

I think we can just decline this. NoteDB works around the problem and the underlying bug leading to this has gone away too.

Every user has SSH access to the server. And you'd need to be a user to do the commit to the repo....who's being excluded here?