|
Consider a patch going into the mainline that has a test which can be run under multiversion, but first requires a patch (perhaps the same, perhaps different) to first be backported/pushed to a release branch
The typical workflow for this is:
- Develop the patch on the latest branch.
- Add an entry to etc/backports_required_for_multiversion_testing.
- Do a patch run – when green:
- Push to mainline
- Cherry pick or develop a patch on the expected release branch
- Do a patch run – when green
- Push to the branch
(Roughly) after that last step a developer does, the mainline branch is eligible to run the test that was previously being skipped over. However, there was no way to test this in a patch run prior to the test being executed as part of a waterfall commit/other people's patch builds. This often creates a lot of testing-only red boxes.
It's not obvious to me we can always safely revert some/all of the patches to get back into a green state.
I couldn't say what a solution looks like, but filing a ticket in hopes of us being able to do something better here.
|