Just came across this mail, thanks for initiating this discussion Kay. To add; another issue which recurs is very rapid commit's: before most contributors have had a chance to even look at the changes proposed. There is not much prior discussion on the jira or pr, and the time between submitting the PR and committing it is < 12 hours.
Particularly relevant when contributors are not on US timezones and/or colocated; I have raised this a few times before when the commit had other side effects not considered. On flip side we have PR's which have been languishing for weeks with little or no activity from committers side - making the contribution stale; so too long a delay is also definitely not the direction to take either ! Regards, Mridul On Tue, Jul 22, 2014 at 2:14 AM, Kay Ousterhout <k...@eecs.berkeley.edu> wrote: > Hi all, > > As the number of committers / contributors on Spark has increased, there > are cases where pull requests get merged before all the review comments > have been addressed. This happens say when one committer points out a > problem with the pull request, and another committer doesn't see the > earlier comment and merges the PR before the comment has been addressed. > This is especially tricky for pull requests with a large number of > comments, because it can be difficult to notice early comments describing > blocking issues. > > This also happens when something accidentally gets merged after the tests > have started but before tests have passed. > > Do folks have ideas on how we can handle this issue? Are there other > projects that have good ways of handling this? It looks like for Hadoop, > people can -1 / +1 on the JIRA. > > -Kay --------------------------------------------------------------------- To unsubscribe, e-mail: dev-unsubscr...@spark.apache.org For additional commands, e-mail: dev-h...@spark.apache.org
