Le 31/10/2026 à 13:15, Eric Covener a écrit :
Hi, vote version of this thread since it looks promising:
https://lists.apache.org/thread.html/dpnfl07kjns2w5q6jlj1r5nppgwkn145
Context (for non-CTR components)
In SVN / STATUS, three committer approvals are required. Self-review
is counted, so in practice there are two independent +1s needed.
I propose with this vote that we reduce this by 1 in practice and also
establish some guidance on what non-STATUS review is.
* With passing (or explainable) CI only two committer approvals are
needed. Without CI, an additional committer review is needed.
* Committer approval includes: github review, PR comment, explicit
dev@ email approval, security@ review, or the merge itself.
This means that in practice for committers Alice and Bob that if Alice
proposes a PR against 2.4.x, and CI doesn't regress, a single review
(including just a LGTM comment) from Bob is all that's needed for
Alice to move on to merging the work.
In the case where Alice is not a committer: Two different committers
should do a combination of passing review/comment/merge. The 2nd
doesn't need to do anything explicit.
[ ] +1: Allow two approvals, including self-approval with CI
[ ] +0: I don't feel great about this but...
[ ] -1: This is bad.
[X] +0: I don't feel great about this but...
but I understand the reason of such a proposal (thanks IA flood...)
I would +1 if the PR updates the test framework with a new test that
fails before the proposal and succeeds after.
CJ
