Hi, The risk of accidental push to wrong remote scared me so much that I developed a git hook with a popup to prevent it. Direct push to apache remote without a PR is so rare for me that I now gate it behind an extra fingerprint authentication.
The setup (Mac) is shared here https://github.com/cominvent/git-push-gate Jan > 8. sep. 2026 kl. 12:10 skrev Chris Hegarty <[email protected]>: > > Hi all, > > I was backporting a change earlier today and made a mistake - I pushed my > local branch_10x to upstream rather than my fork. The implication of this is > that the backport of #16647 is now three commits, rather than a single > squashed commit. Sorry about that. > > The good new is this should be fine in practice. It’s mostly a dirty history > problem, not a broken branch. > > What landed on the branch (three commits on top of what was there before): > * the getVectorCount backport from #16647 > * a small follow-up fox for 10.x APIs > * removal of a empty Faiss file that should not have been there > > I do not have permission to reset the branch, so I left it as-is for now, > rather than pile on reverts. > > I ran a full ./gradle test suit locally on the current tip and it passed. If > you pull branch_10x or CI fails on it or if anything looks off, please ping > me. > > And yeah, it was silly of me to use the branch_10x name as the backport > branch! I typically do not do this. > > -Chris. > --------------------------------------------------------------------- > To unsubscribe, e-mail: [email protected] > For additional commands, e-mail: [email protected] > --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
