Alan,

+1 on branching.

That said, I just filed
https://issues.apache.org/jira/browse/HCATALOG-577 which details a bug
that breaks pig queries using HCatLoader in conjunction with the
distinct operator, followed by HCatStorer. I figure we're potentially
breaking in a number of other ways, but this one is demonstrable.

I think we should definitely have this fixed in this release, but
whether or not we want to do this before we branch, or after we branch
is something I don't mind either way on.

-Sushanth

On Fri, Dec 21, 2012 at 9:05 AM, Alan Gates <[email protected]> wrote:
> I propose we push in the following three JIRAs:
>
>    Add HideUtilityClassConstructor checkstyle rule
>    https://issues.apache.org/jira/browse/HCATALOG-568
>
>    test.sh should let users specify the ant version
>    https://issues.apache.org/jira/browse/HCATALOG-564
>
>    HCATALOG-527 Broke backward compatibility
>    https://issues.apache.org/jira/browse/HCATALOG-576
>
> and then branch for 0.5.  I'd really like to branch now because I am 
> concerned that otherwise we'll keep having new things come in.  Obviously 
> this branch would still be open to any bug fixes.  But the intent would be to 
> not add any new features.
>
> I'll take ownership of pushing in those three patches.
>
> Alan.
>
> On Nov 28, 2012, at 7:57 PM, Travis Crawford wrote:
>
>> I totally agree we should do as much as possible before Hive 0.10 is
>> released, so we can quickly release afterwards. Towards that goal, I
>> spent a bunch of time today reviewing the release process, and seeing
>> what we can do now so we can release quickly when we're ready.
>>
>> I'd love if someone can take a look at
>> https://issues.apache.org/jira/browse/HCATALOG-558 because that makes
>> some critical fixes in our src-release ant target. Additionally, it
>> updates test.sh to run tests from an extracted src-release artifact to
>> prevent fixes like this from being needed in the future.
>>
>> We should update our CI jobs to run test.sh (after updating with the
>> appropriate commands) so we're doing this on a regular basis, as well
>> as producing/archiving the src-release artifact from CI. This will
>> help make the current (and future) releases go very smoothly, as much
>> of the process will be automated.
>>
>> Thanks!
>> Travis
>>
>>
>> On Wed, Nov 28, 2012 at 1:46 PM, Travis Crawford
>> <[email protected]> wrote:
>>> I just hopped in #hcatalog on freenode if anyone wants to discuss/help
>>> with release prep.
>>>
>>> --travis
>>>
>>> On Wed, Nov 28, 2012 at 1:43 PM, Travis Crawford
>>> <[email protected]> wrote:
>>>> Totally agree we can parallelize some things. Here's a list of stuff
>>>> we can do now to help us release faster:
>>>>
>>>> RELEASEAUDIT TARGET
>>>>
>>>> Part of our release process is running the "releaseaudit" target.
>>>> AFAICT the only thing it does is check for appropriate file headers.
>>>> We actually do this with checkstyle now. Is that target still
>>>> necessary? Ideally we could add checkstyle targets for some non-java
>>>> files if necessary so we have an enforcement mechanism that happes
>>>> with each patch, rather than trying to fix this stuff at the last
>>>> minute before releases. If we can finish up this issue that will help
>>>> us get closer to making a release:
>>>>
>>>> releaseaudit target does not cover all code
>>>> https://issues.apache.org/jira/browse/HCATALOG-446
>>>>
>>>>
>>>> BUILD RELEASE ARTIFACT FROM JENKINS
>>>>
>>>> Jenkins has a feature where each build can produce files that are
>>>> archived as long as the build history is around. Currently we use this
>>>> to archive the contents of our "package" target. Can we update the CI
>>>> job configuration to also generate and archive the "src-release"
>>>> artifact? That way, each build will show us what would be released if
>>>> we chose to release at that point. That way, there will be no
>>>> last-minute surprises as to the contents of our release artifact.
>>>>
>>>> Also, we've talked about this in the past, but I think we should
>>>> checkin a script with the CI command so everyone's aware of what tests
>>>> are run. We could update
>>>> http://svn.apache.org/repos/asf/incubator/hcatalog/trunk/build-support/scripts/test.sh
>>>> with the correct commands, then update our CI job to run that script.
>>>>
>>>>
>>>> PUBLISH JAVADOC IN CI
>>>>
>>>> Jenkins has a feature where it can archive and serve javadoc for each
>>>> build. Can we set this up for
>>>> https://builds.apache.org/job/Hcatalog-trunk-build/, which I assume
>>>> we'll clone for the 0.5 branch CI job. This will make viewing javadoc
>>>> really easy so we can see what it looks like before the final release.
>>>> Also, if we do find content/formatting errors it will be easier to
>>>> discuss as we can send links to see what it actually looks like.
>>>>
>>>>
>>>> Actually branching seems like the easy part of making a release. If we
>>>> can get the above items resolved we'll be in good shape to release
>>>> really quickly after Hive releases. I can take care of the
>>>> "releaseaudit" issue after getting more info. Can someone with CI
>>>> permissions take care of producing+archiving the src-release artifact
>>>> and publishing javadoc?
>>>>
>>>> Thanks!
>>>> Travis
>>>>
>>>>
>>>>
>>>>
>>>>
>>>> On Wed, Nov 28, 2012 at 1:11 PM, Alan Gates <[email protected]> wrote:
>>>>> I agree we shouldn't cut an rc before Hive ships.  In my experience 
>>>>> there's usually a week or two of bug fixing between branching and rolling 
>>>>> an rc.  That's what we can parallelize.  If we're confident we can roll 
>>>>> an rc immediately after branching then I'm fine with waiting to branch.  
>>>>> But I do not share this confidence.
>>>>>
>>>>> Alan.
>>>>>
>>>>> On Nov 28, 2012, at 11:26 AM, Travis Crawford wrote:
>>>>>
>>>>>> On Wed, Nov 28, 2012 at 11:10 AM, Alan Gates <[email protected]> 
>>>>>> wrote:
>>>>>>
>>>>>>> We definitely cannot release HCat 0.5 before Hive 0.10.  But I don't 
>>>>>>> think
>>>>>>> we need to wait to branch.  Why not branch now and start fixing issues 
>>>>>>> and
>>>>>>> then update the dependencies once Hive ships?
>>>>>>>
>>>>>>
>>>>>> Doing QA on a moving target seems challenging – I'd prefer to ask people 
>>>>>> to
>>>>>> spend their time testing legitimate release candidates. If we're blocked 
>>>>>> on
>>>>>> making a release candidate until Hive ships, there's no need for a 
>>>>>> release
>>>>>> branch yet. Prematurely branching means we have to track changes in the
>>>>>> release branch & trunk, which is work we can avoid.
>>>>>>
>>>>>> I'm not totally against branching now, but it seems like additional work
>>>>>> for no gain.
>>>>>>
>>>>>> Does anyone have more insight into Hive's release ETA?
>>>>>>
>>>>>> --travis
>>>>>>
>>>>>>
>>>>>> Alan.
>>>>>>>
>>>>>>> On Nov 28, 2012, at 11:03 AM, Travis Crawford wrote:
>>>>>>>
>>>>>>>> Hey hcat gurus -
>>>>>>>>
>>>>>>>> Regarding the 0.5 release, I'd like to postpone branching until Hive
>>>>>>> 0.10.0
>>>>>>>> ships, as we cannot release with snapshot dependencies.
>>>>>>>>
>>>>>>>> Proposal:
>>>>>>>> * Keep working on items with a target fix of 0.5 until Hive ships
>>>>>>>> * Branch when hive 0.10.0 ships, and post a release candidate
>>>>>>>> * Punt everything not yet complete to the next release
>>>>>>>> * Budget ~1 week for people to test, more if issues are discovered
>>>>>>>> * Ship it!
>>>>>>>>
>>>>>>>> Thoughts?
>>>>>>>>
>>>>>>>> --travis
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>> On Wed, Nov 21, 2012 at 2:17 PM, Travis Crawford
>>>>>>>> <[email protected]>wrote:
>>>>>>>>
>>>>>>>>> Status update:
>>>>>>>>>
>>>>>>>>> Thanks all for doing some jira cleanup and finalizing the list of
>>>>>>>>> outstanding issues for 0.5. Open issues scheduled for inclusion are
>>>>>>>>> available here:
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>> https://issues.apache.org/jira/secure/IssueNavigator.jspa?reset=true&jqlQuery=project+%3D+HCATALOG+AND+fixVersion+%3D+%220.5%22+AND+resolution+%3D+Unresolved+ORDER+BY+due+ASC%2C+priority+DESC%2C+created+ASC&mode=hide
>>>>>>>>>
>>>>>>>>> The plan is to branch next Tuesday, and hopefully release by EOW. As
>>>>>>>>> the schedule is tight, please be mindful of the review queue so we can
>>>>>>>>> get through these open issues.
>>>>>>>>>
>>>>>>>>> Thanks!
>>>>>>>>> Travis
>>>>>>>>>
>>>>>>>>>
>>>>>>>>> On Mon, Nov 19, 2012 at 11:21 AM, Travis Crawford
>>>>>>>>> <[email protected]> wrote:
>>>>>>>>>> Also, if we don't already have one I'd like to volunteer as the
>>>>>>>>>> release manager. I've never done one before and would like to learn
>>>>>>>>>> the process.
>>>>>>>>>>
>>>>>>>>>> Thanks!
>>>>>>>>>> Travis
>>>>>>>>>>
>>>>>>>>>> On Mon, Nov 19, 2012 at 10:57 AM, Travis Crawford
>>>>>>>>>> <[email protected]> wrote:
>>>>>>>>>>> I'd love to start thinking about a 0.5 release, but let's hold off 
>>>>>>>>>>> on
>>>>>>>>>>> branching until we understand a bit more about what we're releasing.
>>>>>>>>>>>
>>>>>>>>>>> As a first step, can everyone make sure the "Fix Version/s:" field
>>>>>>>>>>> says 0.5.0 for issues you're hoping to have included?
>>>>>>>>>>>
>>>>>>>>>>> You can view the full list of unresolved issues associated with 
>>>>>>>>>>> 0.5.0
>>>>>>>>> here:
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>
>>>>>>> https://issues.apache.org/jira/secure/IssueNavigator.jspa?reset=true&jqlQuery=project+%3D+HCATALOG+AND+fixVersion+%3D+%220.5%22+AND+resolution+%3D+Unresolved+ORDER+BY+due+ASC%2C+priority+DESC%2C+created+ASC&mode=hide
>>>>>>>>>>>
>>>>>>>>>>> Also, for committers, if you can make some time for reviewing patch
>>>>>>>>>>> available issues that would be much appreciated. Its discouraging 
>>>>>>>>>>> for
>>>>>>>>>>> contributors to have their patches linger.
>>>>>>>>>>>
>>>>>>>>>>> Thanks!
>>>>>>>>>>> Travis
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>> On Mon, Nov 19, 2012 at 10:32 AM, Daniel Dai <[email protected]>
>>>>>>>>> wrote:
>>>>>>>>>>>> Hi, folks,
>>>>>>>>>>>> How about branching 0.5 for 0.5.0 release? Hive will release 0.10.0
>>>>>>>>>>>> soon (http://www.mail-archive.com/[email protected]/msg24708.html
>>>>>>> ),
>>>>>>>>>>>> sounds like it is a good time to make a HCatalog release as well.
>>>>>>>>>>>> Thoughts?
>>>>>>>>>>>>
>>>>>>>>>>>> Thanks,
>>>>>>>>>>>> Daniel
>>>>>>>>>
>>>>>>>
>>>>>>>
>>>>>
>

Reply via email to