+1 for option 1. Seems the only downside of option 1 is that some RC numbers may be non-sequential.
Dongjoon Hyun <dongjoon.h...@gmail.com> 于2025年6月5日周四 07:57写道： > +1 for the proposal, Hyukjin. Thank you for the whole and seamless > migration toward this direction. > > Please make it sure that we explicitly show the human release manager name > and email address (instead of bot sender) in the generated email. That's > the only concern I have. > > Thanks, > Dongjoon. > > > > On Wed, Jun 4, 2025 at 9:32 PM Mridul Muralidharan <mri...@gmail.com> > wrote: > >> >> We can always invalidate the vote with -1 in case it is found to be >> sent incorrectly ... As long as the automation does not end up generating a >> tonne of mails, that is, it should be fairly manageable :) >> I am in favor of automating it with option 1. >> >> Thanks for driving this Hyukjin ! >> >> Regards, >> Mridul >> >> >> On Wed, Jun 4, 2025 at 6:53 PM Hyukjin Kwon <gurwls...@apache.org> wrote: >> >>> Hi all, >>> >>> As some of you may know, I’ve been working on automating the Spark >>> release process (release.yml >>> <https://github.com/apache/spark/actions/workflows/release.yml>). The >>> basic steps are done, and I’m now looking into automating some of the >>> remaining manual tasks. >>> >>> One such task is sending the email to start the vote for an RC. I’d like >>> to automate this step as well. >>> >>> The potential downside is that, in corner cases, an incorrect RC might >>> still trigger the vote email (even though failures should be caught >>> earlier). To handle this, I propose we send the email automatically and >>> rely on the community to help verify the RC. If something is wrong, we can >>> simply cut a new RC - which is now much easier to do. >>> >>> Alternatively, a more conservative option is to generate a draft of the >>> email in the build log and let the release manager copy and send it >>> manually. >>> >>> I personally prefer the first approach, but I’d like to hear what others >>> think. >>> >>>
