Hi Chris, Olivier,
Thanks for the detailed explanation about the interaction with other repositories, I understand the cause of the issue better now.
However, I still feel like changing the strategy with which Jenkins tests the patches would be an elegant and hopefully not too complicated solution. I know that this will cause a patch that actually doesn't rebase cleanly onto the current HEAD at the moment to fail the test, but isn't that a desirable feature? If they don't rebase cleanly that means the author will need to manually intervene before submission anyway, and the CI is supposed to test whether the patch is ready for submission, so I think it makes sense for the CI to fail and notify the author of the problem in that case.
It’s usually not too much hassle, though, to just rebase, given that Gerrit can usually do it automatically with the click of a button. The patches will, inevitably, have to be rebased eventually before merging anyway.
I don't think this applies to the patches currently affected by this problem. They fail because their parent is out of date with other repositories, not because the patch itself would somehow not apply cleanly. If it wasn't for the CI issue these patches should be able to get submitted (i.e. cherry-picked onto HEAD) cleanly by Gerrit without needing another rebase, because the rebase doesn't actually change anything about the patch itself, it just updates the parent.
I know it's just one button click, but I think the problem is more that this issue is hard to understand for new contributors. Before you know to click that button you first have to dig through the CI interface, find the errors, be utterly confused why you're getting build errors in directories completely unrelated to your patch, ask a maintainer for help, and then be told to click the rebase button. If there is an easy automated solution to eliminate all this unnecessary friction, I think it would be nice to implement it.