On 13-02-04 09:10 PM, John Clements wrote: > I'm guessing this was deliberate, but I thought I'd check. The bors bot > (IIUC) looks for an r+ attached to the final commit of a pull request in > order to start auto-merging. When a pull request becomes > non-automatically-mergeable (as mine did a few hours ago: > jbclements:demodeing-and-deGCing), my understanding is that it's a kindness > to force-push to the given branch to make it automatically mergeable (as I > just did). Since the approval message appears on the last commit and the last > commit is renamed, though, that approval is lost. > > Is it the intent of the system to have people re-approve the pull requests in > situations like this?
Yeah. Because bors really has no ability to judge the "mostly-sameness" of a re-submitted chunk of code, if its bits differ. If it didn't work this way, you could get approval on X and then "resubmit" completely-different-code Y and have bors land it. This was, in my estimate, not a desirable behavior for a robot with commit rights. -Graydon _______________________________________________ Rust-dev mailing list [email protected] https://mail.mozilla.org/listinfo/rust-dev
