Thanks for bringing it up, Gopal. Although canceling the patch prevents the spam, it would be convenient if we didn't have to do that. The Apache-wide PreCommit-Admin job is the thing responsible for enqueuing patches to run in our own PreCommit-<project>-Build jobs.
https://builds.apache.org/view/All/job/PreCommit-Admin/ I suspect it would be difficult to change this to match only on *.patch, because it would require coordination across a lot of Apache projects to approve the change. Another possible solution would be to do an early exit from our own test-patch.sh if the file doesn't match *.patch. We'd still end up with job executions showing up in Jenkins, but at least they wouldn't comment on the jiras. I think it's an idea worth exploring. Let's go ahead and file a jira for it, unless others have objections. Chris Nauroth Hortonworks http://hortonworks.com/ On 4/8/15, 8:32 AM, "Allen Wittenauer" <a...@altiscale.com> wrote: > >Cancel the patch and it won’t do that. > >Also, FWIW, HADOOP-11746 makes test-patch.sh run significantly faster for >non-Java patches. (of course, the current test-patch doesn’t publish >times, so it’s hard to know how slow it is. lol) > >On Apr 8, 2015, at 12:39 AM, Gopal Vijayaraghavan <gop...@apache.org> >wrote: > >> Hi, >> >> I keep uploading logs and screenshots to HADOOP bugs which trigger unit >> test-runs, which fail because a .log.gz or .png file isn¹t really going >>to >> work. >> >> For instance, https://builds.apache.org/job/PreCommit-HADOOP-Build/6074/ >> has a PNG image in the patch build artifact. >> >> >> Can the hadoop-qa jenkins job filter on the extension of the attached >> files instead of spamming JIRA with false-positive failure markers? >> >> Cheers, >> Gopal >> >> >
