At the Hive contributor meeting, we discussed this and came to the conclusion
that the failures reported so far are ignorable based on the Hudson history
(and in this case Ning's observation of JVM inconsistencies with respect to
serialization format).
We need one more +1 from a committer before we can release.
JVS
On Oct 25, 2010, at 12:54 PM, Ashish Thusoo wrote:
> I got the following test failures on the release candidate...
>
> groupby2.q
> groupby3.q
> groupby4.q
> groupby5.q
> groupby6.q
>
> not sure if this is just in my env or if others have seen this...
>
> A sample of the diff is below and seems to be related to some plan ordering
> or some change in plan. Is anyone else getting this?
>
> Ashish
>
> -------------------------
> [junit] diff -b -I'\(\(<java version=".*"
> class="java.beans.XMLDecoder">\)\|\(<string>.*/tmp/.*</string>\)\|\(<string>file:.*</string>\)\|\(<string>[0-9]\{10\}</string>\)\|\(<string>/.*/warehouse/.*</string>\)\)'
>
> /data/users/athusoo/tmp/hive-0.6.0/src/build/ql/test/logs/positive/groupby6.q.xml
>
> /data/users/athusoo/tmp/hive-0.6.0/src/ql/src/test/results/compiler/plan/groupby6.q.xml
> [junit] 352,353c352
> [junit] < <object class="java.lang.Enum"
> method="valueOf">
> [junit] <
> <class>org.apache.hadoop.hive.ql.plan.GroupByDesc$Mode</class>
> [junit] ---
> [junit] > <object
> class="org.apache.hadoop.hive.ql.plan.GroupByDesc$Mode" method="valueOf">
> [junit] 878,879c877
> [junit] < <object class="java.lang.Enum" method="valueOf">
> [junit] <
> <class>org.apache.hadoop.hive.ql.plan.GroupByDesc$Mode</class>
> [junit] ---
> [junit] > <object
> class="org.apache.hadoop.hive.ql.plan.GroupByDesc$Mode" method="valueOf">
>
> --------------------------
>
> ________________________________________
> From: John Sichi [jsi...@facebook.com]
> Sent: Thursday, October 21, 2010 12:22 PM
> To: <dev@hive.apache.org>
> Subject: Re: [VOTE] hive 0.6.0 release candidate 0
>
> Yeah, the scripts should only be needed in configurations where JDO is told
> not to automatically update the schema. This is recommended for production
> environments.
>
> For this particular release, taking a downtime while running the scripts is a
> good idea due to the nature of the changes (e.g. altering the primary key on
> COLS). That needn't be true in general for additive-only changes.
>
> JVS
>
> On Oct 21, 2010, at 12:14 PM, Edward Capriolo wrote:
>
>> On Wed, Oct 20, 2010 at 6:38 PM, John Sichi <jsi...@facebook.com> wrote:
>>> The tarballs are at
>>>
>>> http://people.apache.org/~jvs/hive-0.6.0-candidate-0
>>>
>>> Carl did some sanity testing on it already, but any additional testing you
>>> can do before voting helps to ensure a quality release.
>>>
>>> JVS
>>>
>>>
>>
>> I am checking it out now. It seems like since i have used two trunk
>> versions since hive the view related tables have already been created.
>> I do not need the update script.
>
