Awesome, thank you very much Brock! Jarcec
On Dec 19, 2012, at 10:38 PM, Brock Noland <[email protected]> wrote: > Just an update here, the last two comments on this JIRA > https://issues.apache.org/jira/browse/FLUME-1787 are the messages on > failure and success. At present the only check it does is apply the > patch and run the tests. We can add more checks over time. I ended up > not using the Hadoop script because it is one ugly beast of a shell > script. The patches on that jira are only for testing this script. > I'll be submitting a real patch soon. > > Brock > > On Sat, Dec 8, 2012 at 2:39 AM, Mike Percy <[email protected]> wrote: >> I really like the idea of doing this! +1 from me if we can swing it. >> >> Regards, >> Mike >> >> >> On Fri, Dec 7, 2012 at 3:20 PM, Jarek Jarcec Cecho <[email protected]>wrote: >> >>> Btw I just found very interesting draft that is explaining quite huge >>> portions of the design: >>> >>> http://wiki.apache.org/general/PreCommitBuilds >>> >>> Jarcec >>> >>> On Fri, Dec 07, 2012 at 11:14:13AM -0600, Brock Noland wrote: >>>> Hi, >>>> >>>> Great to hear you were thinking the same thing! I have found the >>>> script they use to do the checking: >>>> >>>> >>> https://github.com/apache/hadoop-common/blob/trunk/dev-support/test-patch.sh >>>> >>>> and I think we could probably implement something similar. I agree >>>> that our build machines are much busier so I don't think we can use it >>>> as a hard requirement for commit. The biggest open question in my mind >>>> is how do they query JIRA for patches? Grab the correct patch off of >>>> the JIRA? Schedule the job to poll for these things? >>>> >>>> Brock >>>> >>>> On Fri, Dec 7, 2012 at 10:26 AM, Jarek Jarcec Cecho <[email protected]> >>> wrote: >>>>> Hi Brock, >>>>> that is an excellent idea and I'm definitely in favour of that. In >>> fact I wanted to propose something similar myself. >>>>> >>>>> However I did not quite finished my own investigation yet. It seems >>> that this pre commit hooks are done on projects that were mostly donated by >>> Yahoo! (Hadoop, Oozie) on Jenkins nodes named "hadoopX". The hadoop named >>> nodes have comment "Build machine for Hadoop project builds", so I'm not >>> sure that they are available for other projects as well. >>>>> >>>>> With removing those 9 hadoop notes, it's like 10 other nodes in total >>> for entire Apache organization and I do know that sometimes the build queue >>> might be quite huge. Having said that it might be good idea to ask Infra >>> team if pre-commit hooks are available to all projects or only to projects >>> that have "special hardware" available. >>>>> >>>>> Jarcec >>>>> >>>>> On Fri, Dec 07, 2012 at 10:11:53AM -0600, Brock Noland wrote: >>>>>> Hi, >>>>>> >>>>>> I like how in the Hadoop project if you mark a patch "Patch Available" >>>>>> jenkins applies the patch to trunk, runs the tests, and outputs the >>>>>> result to the JIRA. Here is one example: >>>>>> >>>>>> http://s.apache.org/qz >>>>>> >>>>>> I think we should still run tests manually before commit. However, >>>>>> this will often give a reviewer an indicator of whether the change >>>>>> breaks a test without even having to download the code. >>>>>> >>>>>> Thoughts? >>>>>> >>>>>> Brock >>>> >>>> >>>> >>>> -- >>>> Apache MRUnit - Unit testing MapReduce - >>> http://incubator.apache.org/mrunit/ >>> > > > > -- > Apache MRUnit - Unit testing MapReduce - http://incubator.apache.org/mrunit/
