incubator-bigtop-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Andrew Bayer <andrew.ba...@gmail.com>
Subject Re: iTest.next
Date Mon, 07 Nov 2011 21:32:45 GMT
As the guy who suggested Gradle, let me justify that suggestion. =)

I love Maven. I mean, I *love* Maven. It's the best pure build tool for
Java, period. But it's a specialized tool - it's a hammer, and builds are
nails. The Bigtop test use case is not a nail - while we definitely need a
lot of thing Maven provides (test execution+reporting and dependency
management most notably), we're not doing actual "builds" in our test
executions, we're diverging from the lifecycle, we'd like to have more
granular control over which tests are getting executed, etc, etc... Hacking
Maven plugins is not fun at all, so we really don't want to go down the
route of having to repurpose/extend/modify/write our own Maven plugins.

Gradle, on the other hand, is much more flexible. It's closer to Ant than
Maven in terms of mandatory convention over configuration, and frankly, I'd
much rather script in Groovy than in XML. I'm not suggesting we should use
Gradle for the entire Hadoop ecosystem, or even necessarily the Bigtop test
artifact builds - I just believe that, for test execution, Gradle is a more
appropriate tool than Maven, and can give us more flexibility and
extensibility in the areas we care about.

A.

On Mon, Nov 7, 2011 at 1:24 PM, Matt Foley <mfoley@hortonworks.com> wrote:

> Looking at Gradle, I found this disconcerting statement (
>
> http://stackoverflow.com/questions/1163173/why-use-gradle-instead-of-ant-or-maven
> ):
>
> > Groovy is much nicer to code than XML.
>
>
> Are you proposing that we manage the build system in Groovy?
> Unless the whole ecosystem can be moved in this direction, I would
> recommend against it.
>
> --Matt
>
> On Mon, Nov 7, 2011 at 11:57 AM, Tom White <tom@cloudera.com> wrote:
>
> > +1 to lowering the barrier to entry for contributing new tests (#7).
> >
> > Also, rather than introduce a new build system, why not improve the
> > plugin (failsafe) if it has deficiencies?
> >
> > Tom
> >
> > On Fri, Nov 4, 2011 at 4:00 PM, Roman Shaposhnik <rvs@apache.org> wrote:
> > > Folks,
> > >
> > > one of the biggest challenges that I'd like to tackle post Bigtop
> > > 0.2.0 is increasing
> > > the scope, coverage and robustness of the iTest and its test artifacts.
> > To that
> > > end I've started a Wiki page outlining the issues:
> > >    https://cwiki.apache.org/confluence/display/BIGTOP/iTest.next
> > >
> > > Feel free to contribute and I really hope we can spend some time at
> > ApacheCON
> > > talking about that.
> > >
> > > Thanks,
> > > Roman.
> > >
> >
>

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