On Fri, Mar 03, 2023 at 02:30:40PM +1100, Ian Wienand wrote:
The changes to update copy-conditionals and submit-requirements have multiplied a bit, so I thought it might be good to have a bit of a written-down plan to get us ready for Gerrit 3.7.
Thanks for the feedback (particularly clarkb) so here's a small update just to get things straight The changes to All-Projects have fixed an issue with the copyConditional matching NO_CHANGE (which would allow updating commit messages, which we don't want). https://review.opendev.org/c/opendev/system-config/+/876236 https://review.opendev.org/c/opendev/system-config/+/876237 Those are incoporated into an updated patch for the production All-Projects at https://paste.opendev.org/show/brAj40R1mJbQZSXAXEQ5/ I have tested this on a held node https://104.130.253.50/c/x/test-project/+/3 it's going to look something like https://imgur.com/a/7aWTHix with the code-review, workflow and verfied as submit-requirements, with "trigger" requirements like "Review-Priority" listed separately. There was some discussion about when it was OK to push this. Personally I think probably when we are ready, but with us watching for a quick revert if required -- the idea is that for users it should be a no-op; so the rollback condition would be anything not working as it does before the update. I was thinking that because the long-term implications of "function = NoOp" being deprecated are unclear, we should use a submit-requirement for our no-op labels like Backport-Candidate, etc. Upon playing with this more, I think that's not a good idea. You can do this, with a submit-requirement that "is:true" so it always passes. However, it looks confusing in the UI. If you have a "-1 Not a candidate", then you get a "passing" submit-requirement for the Backport-Candidate with a -1. Its too confusing to deal with. It is unclear how this sort of thing will be handled into the future by gerrit. The "function" is deprecated, but if it will be completely removed we don't know; see https://groups.google.com/g/repo-discuss/c/tZwIp3Hx-wA/m/rJgmT_OsAQAJ Thus I've abandonded https://review.opendev.org/c/openstack/project-config/+/875804 and simlified to just convert everything to NoOp with https://review.opendev.org/c/openstack/project-config/+/875993 At least then we are only dealing with one type of function in our ACL's consistently ... The ACL's that *are* using blocking submit rules have had a couple of things pointed out by clarkb fixed, and are converted with https://review.opendev.org/c/openstack/project-config/+/875995 https://review.opendev.org/c/openstack/project-config/+/875996 I think this series of patches puts us in a good position for a 3.7 upgrade. -i