www-jcp-open mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Geir Magnusson Jr." <g...@pobox.com>
Subject Re: TSFKAJ
Date Sat, 14 Jul 2007 18:03:27 GMT

On Jul 13, 2007, at 8:06 AM, Mark Wielaard wrote:

> Hi Steve,
> Steve Loughran <stevel@...> writes:
>> Mark Wielaard wrote:
>>> http://www.infoq.com/news/2007/06/openjdk-hybrids
>> Interesting. Now, if you want to keep maintenance costs down you  
>> want to
>> get your fixes back in to the main branch, and no doubt sun would  
>> like
>> to take on the bugfixes that get proxy support working on kde  
>> desktops,
>> etc. But nobody likes to take patches from random developers. Patches
>> that say "Here's my fix to the proxy code, it passes all tests"  
>> you'd be
>> much happier. Having a completely open test suite makes it much  
>> easier,
>> especially if it comes with a set of VMware images to emulate  
>> reference
>> platforms (or the OSS ones, anyway)
>> Looking at the IcedTea release, it appears they are stubbing out the
>> stuff that is restricted: sound, snmp and bits of AWT.
> And some crypto stuff. But that was just phase 1. Getting a  
> completely free
> setup without any proprietary encumbrances buildable and  
> bootstrappable with
> only free software on GNU/Linux systems. Which is what IcedTea 1.0  
> was.

I think this is cool.  It's a shame Sun didn't do this.

> For 1.1 (hopefully out today or this weekend) the stubbed out bits  
> are slowly
> replaced with full free versions. Crypto (and in particular https  
> ssl support)
> is now done http://mail.openjdk.java.net/pipermail/discuss/2007- 
> July/000103.html
> for graphics various efforts are underway either replacing it  
> wholesale with the
> GNU Classpath awt peers
> http://kennke.org/blog/2007/07/11/swing-on-gtk-peers-again/ or  
> replacing the
> encumbered low level graphics and font rasterizers with free versions
> http://fkung.wordpress.com/2007/06/25/icedtea-graphics/ and
> http://fkung.wordpress.com/2007/07/06/more-fonts/ there is also a  
> summary of the
> efforts on java.net
> http://weblogs.java.net/blog/editors/archives/2007/07/ 
> just_out_of_rea.html

This is cool too.  Do you have any plan to re-sync w/ openJPA?

>> For server-side
>> JREs, you probably dont need a functional sound subsystem, and its
>> pretty hard to test too -I suspect that the TCK doesn't, not  
>> unless it
>> requires you to connect audio out to audio in so that it can verify
>> round tripping, and even then you can't easily distinguish between a
>> loopback in software, a noisy loopback in software and real sound
>> hardware. AWT is testable, and the TCK may have something there,  
>> but you
>> are up against the problem of verifying acceptability of different
>> rendering algorithms. Bit-for-bit image comparision works for  
>> regression
>> testing on identical hardware, but not for testing different  
>> renderers,
>> which, on windows, can mean different display drivers.
> Yes, testing is very, very hard, especially for the things you  
> mention. But
> there is actually a free testsuite available, Mauve
> http://www.sourceware.org/mauve/ which is already used by gcj, gnu  
> classpath,
> kaffe, ikvm, harmony, etc. This of course isn't the official TCK  
> but it is a
> very nice thing to have anyway for all the reasons you give.  
> Interestingly
> enough now that the java reference implementation is available  
> under the GPL
> people have started using it more with the official JDK and  
> actually found
> various bugs with it which are now filed and being worked on for  
> openjdk.
> Although to be fair, there were a lot more bugs found in the mauve  
> testsuite
> itself than there were bugs found in the reference implementation.  
> But it is a
> nice testsuite and a good alternative to the currently proprietary  
> JCK.
> BTW. For sound we do actually have a google summer of code project
> http://code.google.com/soc/2007/gnu/appinfo.html? 
> csaid=FDB6DFFB07FE351C the
> student doing that is also creating Mauve tests to verify his work.
> So if no solution is found for the JCK (openjdk also doesn't have  
> access to it
> yet, GNU Classpath, kaffe, gcj have asked years ago but never got  
> through all
> the red tap - this was one of the reasons for starting harmony btw  
> since we
> hoped Apache would have a lot more experience with that - although  
> accepting
> NDAs was something most GNU hackers would never agree to)

Serious question - does the FSF have any private lists?

> then it might be a
> good idea to bundle forces and push Mauve forward as the shared  
> testsuite to use
> for all the TSFKAJs.

Yes - also, in Harmony, we have a lot of good tests too. (I remember  
you guys using the GUI tests that Intel contributed.)

We should find a way to make it easy to bring together.  Some sort of  
packaging convention or something...


> Cheers,
> Mark

View raw message