Will Webstart be integrated in OpenJDK?

Mark Wielaard mark at klomp.org
Thu Sep 11 10:18:49 UTC 2008

On Thu, 2008-09-11 at 10:24 +0100, Andrew Haley wrote:
> Andrew John Hughes wrote:
> > The biggest problem at the moment is the JCK.  I'm quite happy for
> > Sun to use this to certify their own mashed-up proprietary/Free
> > builds inside Sun labs, and equally it's been great to see this
> > applied to a purely Free build by Red Hat.  However, what disturbed
> > me recently was seeing this used as a tool for patch approval during
> > the AWT/Swing/Java2D discussion Mario mentions.
> Why not?  A gcc patch that failed Plum Hall testing would be rejected
> too, assuming (or course) that the Plum Hall test was valid.

I couldn't find any recent examples of Plum Hall testing against gcc
patch reviews and rejection of patches because of them. But I am pretty
sure the submitter isn't responsible for getting a license agreement
with Plum Hall for contributing to GCC.

The problem with the TCK is precisely that you need to enter an NDA
agreement with Sun over it and that there is no public discussion about
the validity of the tests. Publicly it isn't even know which patches
went in because they made a TCK test pass or because they just
invalidated a TCK test and got it added to the exception lists.

The problem with the current NDA TCK setup is that that you cannot share
tests and code snippets from it with the rest of the community, your
users and customers. You either end up rewriting the tests so you can
publicly share it with others on the mailinglists. Or you have to say
"just trust me, the TCK tests for this particular corner case in this
particular way, so this patch is necessary even though I cannot really
proof it". Neither is really satisfactory. But I would opt for the
rewriting the tests so we have a free replacement.

> > A Free project being dependent on a proprietary test suite for
> > patches is just as bad as it being dependent on proprietary tools to
> > build.  As Mark mentioned in reply to this, we should work towards
> > improving jtreg and Mauve to ensure a Free test suite is available
> > and not rely on the JCK.
> It's not going to happen.  The TCK tests a whole lot of minute details
> of Java langauge compatibility that are not fully explained in the
> Javadoc, and so any open test suite that does not derive from the TCK
> will not be complete.

That could be true, but if so, it is even more important to get this TCK
issue resoled and get a free replacement. It clearly points out a
deficiency in our current documentation that you are unable to tell from
them what the details of a particular method or class really are.

I think having a free TCK should be one of the goals for JDK7 at least.



More information about the discuss mailing list