Yup, I think you expressed it very well, Jay!

Cos

On Mon, Feb 09, 2015 at 09:31AM, jay vyas wrote:
> Okay, so if we do package tests, maybe to make it less work,  we can make a
> generic RPM and DEB packager which  require no maintainance i.e.
> -  just copies gradlew into /usr/share/bigtop-smokes-tests
> - copies tests/ dir into ^ (as described in
> https://issues.apache.org/jira/browse/BIGTOP-1608) .
> 
> After all, there are no RPM or DEB specific features to the tests, and
> lovely gradlew handles all the dependencies.
> 
> 
> On Mon, Feb 9, 2015 at 3:17 AM, Martin Bukatovic <[email protected]>
> wrote:
> 
> > On 02/09/2015 06:09 AM, Konstantin Boudnik wrote:
> > > Yup. We consider tests to tbe first class citizen, but aren't distribute
> > them
> > > as such.
> > >
> > > @Jay: hacking is great and all that. My point is different: tests should
> > be
> > > fixed at the point of the release (and they are); but there's no way to
> > run
> > > them easily against an installed cluster, unless you clone the workspace
> > and
> > > checkout the correct tag. Which isn't very user-friendly.
> >
> > I agree. If you would like to check your Hadoop deployment quickly,
> > tests packaged as part of the release would make it very convenient.
> > It wouldn't make hacking on the tests any harder than it's now, but
> > it would open new use cases and make tests more reusable.
> >
> > --
> > Martin Bukatovic
> >
> 
> 
> 
> -- 
> jay vyas

Reply via email to