On Thu, Sep 5, 2019 at 7:19 AM Steve Loughran <ste...@cloudera.com> wrote: > > > > On Wed, Sep 4, 2019 at 11:11 PM Christopher <ctubb...@apache.org> wrote: >> >> On Wed, Sep 4, 2019 at 3:16 PM Sean Busbey <bus...@cloudera.com> wrote: >> > >> > On Wed, Sep 4, 2019 at 1:35 PM Christopher <ctubb...@apache.org> wrote: >> > > >> > > Hi Hadoop Devs, >> > > >> > > Just curious whether we can get some pull requests merged in. Yetus >> > > will continue to spam open pull requests with walls of text, drowning >> > > out any potential human conversation. This doesn't seem to be >> > > particularly helpful in general, but it is especially bad if the pull >> > > request is open for a long time, and isn't addressed. >> > > >> > >> > Just as a point of clarification, those automated checks are testing >> > something new each time. Specifically how the PR works with the >> > then-current version of the target branch. They also only do so once a >> > week. >> >> Understood. However, even one of those massive messages is too spammy. >> And for a PR open since January, that's 26 such messages. The issue >> isn't that they aren't useful to check that it works against the >> latest... I can see how that's useful. Rather, the issue is that it >> drowns out humans by overwhelming the thread with information that is >> not helpful to a contributor. > > > > yes, it's noisy. I tend to delete all but the last one from the PRs.
Unfortunately, that's only an option for committers. > > The main problem I have is that it means the github notifications window is > full of yetus reports for old patches, which hide real reviews. I did discover that you can block the yetus GitHub user, and that will stop the notifications, but it won't hide their comments on the issues. And, it won't help if the user changes (which has happened at least once since January). >> >> >> > >> > > >> > > 1. Why is there no active response to users that submit pull requests? >> > >> > As with most Open Source projects, and especially ASF projects, my >> > experience in Hadoop is that the easiest way to see more things >> > reviewed and merged that you care about is to review the work of >> > others in order to help the overall reviewer bandwidth. >> >> This is a good suggestion. However, my (now stale) review occurred 2 >> days after it was submitted back in January, wherein I made a small >> suggestion for improvement. Since I'm not a Hadoop committer, I don't >> think there's much more I could have done. Since there was no response >> from anybody within the Hadoop community itself (either concurring >> with my suggestion, or otherwise), and since nobody responded to the >> contributor, my review never mattered. >> >> > >> > > 2. Does Yetus really need to drown human conversation threads on >> > > GitHub with walls of automated text? (vs. using the GitHub checks API >> > > or similar brief status message) >> > >> > Yetus itself can be run in a multitude of ways. The behavior you see >> > is specific to how the Hadoop project runs it. >> >> Interesting. I would recommend it stop being run that specific way... >> since it's unfriendly to contributors, drowns out human conversation, >> and probably triggers a lot of spam on the notifications lists and any >> JIRA issues mentioned, making it more likely to be ignored there as >> well. It is a nice concept, but I don't think it works well on GitHub >> comment threads. (it probably works best on JIRA's "Work Log", and (if >> possible) using GitHub's status/checks API for PRs. If GitHub had a >> "work log" like JIRA, I would recommend it publish to that channel >> instead, but that doesn't currently exist. >> > > kicking off when explicitly asked would be the best; or just give up on PRs > > X months old. Yes, that includes many of my spare-time-but-relevant PRs...but > I know they generally need rebase and merge conflict resolution after a few > months I like the idea of explicitly asking... or better yet, running the tasks in Jenkins or another external CI service (standalone app?) instead of as a GitHub bot. I would have thought Jenkins was configured for Hadoop to do some CI checks on PRs already, but perhaps not? --------------------------------------------------------------------- To unsubscribe, e-mail: general-unsubscr...@hadoop.apache.org For additional commands, e-mail: general-h...@hadoop.apache.org
