Automate the merge of the weekly branch cut
Closed, ResolvedPublic
Actions

Description

This would 'just' mean adding %vote=Code-Review+2 to the git ref, I think. A quick patch in the branch.py code to add a new flag (default False) and then a fiddle on the jenkins job.

Details

	Subject	Repo	Branch	Lines +/-
	make-release: Add `--push-option` to branch.py	mediawiki/tools/release	master	+13 -4

Customize query in gerrit

Related Objects
Search...

		Status	Subtype	Assigned	Task
		Resolved		dancy	T196515 Automate the Train
		Resolved		• dduvall	T278993 Automate the merge of the weekly branch cut

Event Timeline

Jdforrester-WMF created this task.Mar 31 2021, 6:05 PM

Jdforrester-WMF mentioned this in T277507: Core subproject commits aren't updated for backports if wmf branch commit has not yet been merged..Apr 1 2021, 11:46 PM

• LarsWirzenius subscribed.Apr 6 2021, 4:07 PM

May be controversial, added to discussion at next Release-Engineering-Team team meeting

thcipriani moved this task from INBOX to New Work on the Release-Engineering-Team-TODO (2021-04-01 to 2021-06-30 (Q4)) board.Apr 6 2021, 4:09 PM

If we're auto +2'ing, does the branch cut change still need to go through review? I ask because branch.py already has a --no-review option which could skip that process entirely.

In T278993#6977249, @dduvall wrote:

If we're auto +2'ing, does the branch cut change still need to go through review? I ask because branch.py already has a --no-review option which could skip that process entirely.

It's possible that the branch would fail CI for some reason; I think we'd want to know that ahead of time, rather than when we make the first back-port. --no-review bypasses CI and just force-pushes, I think?

In T278993#6977267, @Jdforrester-WMF wrote:

In T278993#6977249, @dduvall wrote:

If we're auto +2'ing, does the branch cut change still need to go through review? I ask because branch.py already has a --no-review option which could skip that process entirely.

It's possible that the branch would fail CI for some reason; I think we'd want to know that ahead of time, rather than when we make the first back-port. --no-review bypasses CI and just force-pushes, I think?