Github user srowen commented on the issue: https://github.com/apache/spark/pull/21060 This certainly looks like a bug fix. I don't know this area well, but I don't see an argument here that the current behavior is correct. Right? When we say we don't back-port behavior changes, we mean "changes in what is meant to be correct behavior". All bug fixes change behavior, but to restore correct behavior. So I don't see an argument against back-porting because it's a behavior change. Of course, sometimes practical concerns override that. If we thought programs were relying on the 'wrong' behavior then we'd have to think twice about correcting it. I don't see that argument being made here, but, I'm not sure? There is evidence the 'wrong' behavior is impacting users though? @gatorsmile I must say I don't understand your position here, can you clarify? So far standard practice here says this is a reasonable backport. What's different here?
--- --------------------------------------------------------------------- To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org For additional commands, e-mail: reviews-h...@spark.apache.org