orc-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Owen O'Malley" <owen.omal...@gmail.com>
Subject Re: [RESULT][VOTE] Shall we release ORC 1.5.6rc1?
Date Thu, 27 Jun 2019 23:25:03 GMT
Thank you, Dongjoon.

I took sometime this morning and didn't see what was going wrong. I think
for right now, I'll finalize this release and we can roll a follow up when
we find and fix the problem. I believe the 1.5.6rc1 is actually better for
most users than 1.5.5.

On Wed, Jun 26, 2019 at 10:06 PM Dongjoon Hyun <dongjoon.hyun@gmail.com>
wrote:

> Thank you, Owen.
>
> Yes, I know that ORC-498 resolved the two NN operations issue and the code
> is directly related to this area.
>
> First of all, I filed https://issues.apache.org/jira/browse/ORC-525 for
> the
> observed situation as a `Major` issue (not a blocker).
>
> I built/installed/tested new ORC with OracleJDK8/AdoptOpenJDK8 on two
> MacBooks multiple times.
> (and AdoptOpenJDK11 additionally)
>
> The failure consistently happens with Apache Spark master(3.0.0) at
> `SQLQuerySuite`.
>
> Interestingly, the corresponding test suite in Apache Spark
> branch-2.4(2.4.4) doesn't fail consistently.
>
> Instead, ORC 1.5.6 might reopen the flaky issue (SPARK-26427: closed with
> ORC 1.5.4) instead at `branch-2.4`.
>
> Bests,
> Dongjoon.
>
>
> On Wed, Jun 26, 2019 at 9:05 PM Owen O'Malley <owen.omalley@gmail.com>
> wrote:
>
> > Dongjoon,
> >   If you need more time, I can hold off. For resource leakage, I'd be
> > worried about ORC-498. I tried to make sure things didn't leak, but that
> > was the change that was sharing a file handle between the ReaderImpl and
> > the RecordReaderImpl so that we didn't end up with two NN operations when
> > reading a file.
> >
> > .. Owen
> >
> > On Wed, Jun 26, 2019 at 8:51 PM Dongjoon Hyun <dongjoon.hyun@gmail.com>
> > wrote:
> >
> > > Hi, All.
> > >
> > > Sorry, but Apache ORC 1.5.6-rc1 seems to have a regression.
> > > Unlike Apache ORC 1.5.5, resource leakage are observed during Apache
> > Spark
> > > sql module unit tests.
> > >
> > > I've been spending time to retest this to make it sure.
> > > Could you give me more time?
> > >
> > > Bests,
> > > Dongjoon.
> > >
> > > On Wed, Jun 26, 2019 at 8:50 PM Owen O'Malley <owen.omalley@gmail.com>
> > > wrote:
> > >
> > > > With three +1's the vote passes. I'll finalize the release. Thanks,
> > Alan
> > > > and Deepak!
> > > >
> > > > .. Owen
> > > >
> > > > > On Jun 26, 2019, at 20:25, Deepak Majeti <majeti.deepak@gmail.com>
> > > > wrote:
> > > > >
> > > > > +1
> > > > > - checked checksum and signature
> > > > > - built from tar in release mode
> > > > > - ran unit tests
> > > > > - ran rat checks
> > > > > - was able to generate a package
> > > > >
> > > > >> On Wed, Jun 26, 2019 at 5:16 PM Alan Gates <alanfgates@gmail.com>
> > > > wrote:
> > > > >>
> > > > >> +1.  Checked the LICENSE and NOTICE files, did a build, ran the
> unit
> > > > >> tests.  On the Java side did a build with a clean repo and ran
> RAT.
> > > (I
> > > > >> don't know how to run RAT on the C++ side but it would be good
to
> > > > somehow
> > > > >> make the same license checks.)
> > > > >>
> > > > >> One small nit, the copyright date in NOTICE is way out of date
> > (2015).
> > > > I
> > > > >> already updated it in both master and branch-1.5 so it should
be
> > fixed
> > > > for
> > > > >> future releases.
> > > > >>
> > > > >> Alan.
> > > > >>
> > > > >> On Fri, Jun 21, 2019 at 9:31 PM Owen O'Malley <
> > owen.omalley@gmail.com
> > > >
> > > > >> wrote:
> > > > >>
> > > > >>> All,
> > > > >>>
> > > > >>> Should we release the following artifacts as ORC 1.5.6?
> > > > >>>
> > > > >>> tar: http://home.apache.org/~omalley/orc-1.5.6/
> > > > >>> tag: https://github.com/apache/orc/releases/tag/release-1.5.6rc1
> > > > >>> jiras:
> > https://issues.apache.org/jira/projects/ORC/versions/12345170
> > > > >>>
> > > > >>> Thanks!
> > > > >>>
> > > > >>
> > > > >
> > > > >
> > > > > --
> > > > > regards,
> > > > > Deepak Majeti
> > > >
> > >
> >
>

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message