Repository navigation
Conversation
|
Review requested:
|
e529655 to
3050fa8
Compare
|
I'd prefer we align it with #66503, for consistency. FWIW force pushing does not invalidate previous approvals nor CI, but yeah someone needs to force push back the previous commit for the CQ to detect it |
|
I believe non-bot commit presence being a skip condition is orthogonal to what this change is for. It can come too, but isn't resolving what I'm trying to resolve. And I think it's solved here too as a side effect of not pushing over the same version update. |
aduh95
left a comment
There was a problem hiding this comment.
I don't think it's the right approach, we should fetch the PR_BRANCH first, apply it's changes before running the update-openssl.sh script
3050fa8 to
a81cea7
Compare
|
Updated to replay the existing update branch before running the script, so its version check handles the no-op. And it also keeps non-bot commits as they were. Manual dispatch bypasses the replay to allow regeneration. |
aduh95
left a comment
There was a problem hiding this comment.
I don't like the fetch-depth: 0, that's really wasteful. Also, this would pick up commits from the PR_BRANCH even if there's no open PR for it: in case the PR was closed but the ref wasn't deleted, we should ignore it.
I think we can solve both problems by making an API call:
a81cea7 to
729e83f
Compare
|
Updated to use the API. I used raw SHAs in an array because |
Replay the existing update branch onto the current target branch before running the updater on scheduled runs. Its version check then skips the same version without rewriting the PR, preserving approvals and CI. This also carries contributor commits into a newer update. Manual dispatches bypass the replay so the update can still be regenerated. Signed-off-by: Filip Skokan <panva.ip@gmail.com> Assisted-by: Codex
729e83f to
7f06241
Compare
|
Why the dont-land labels? Don't we want to backport this? |
It'll need manual backport to 26.x like the previous fix to version sorting does. I have that in my backlog once 26.x-staging gets #66704, or ad-hoc if 3.5.10 releases before that (that will currently fail to get picked up). |
|
Switched to backport-blocked instead |
Replay the existing update branch onto the current target branch before running the updater on scheduled runs. Its version check then skips the same version without rewriting the PR, preserving approvals and CI.
This also carries non-bot commits into a newer update. Manual dispatches bypass the replay so the update can still be regenerated.