Re: libgcj-classpath-20060719 merge

[Thomas Fitzsimmons <fitzsim at redhat dot com>]

Hi Mark,

Mark Wielaard wrote:
> Hi Tom,
>
> On Thu, 2006-07-20 at 08:56 -0400, Thomas Fitzsimmons wrote:
> >   Not fully tested and
> > > not ChangeLog yet. But Hello World works, Eclipse 3.2 starts up and the
> > > Free Swing Demo looks nice.
> > What architecture did you test on?
>
> x86-32 FC5 + some random FC development updates (kernel, gcc, libc,
> eclipse, etc). Also ran make check-target-libjava while sleeping.
> Everything passes.
>
> > > What more do you want anyway!
> > Can you also test gappletviewer and gcjwebplugin?
>
> I can try them out tonight. Anything special to do for
> installing/configuring gcjwebplugin?

Yes, you must add --enable-plugin to the GCC configure line.  Then to 
install it you must create a link from 
~/.mozilla/plugins/libgcjwebplugin.so to 
$prefix/lib*/gcj-4.2.0/libgcjwebplugin.so.

 Any special applets you think are
> "must work"?

No, just try a sampling of applets at

http://developer.classpath.org/mediation/Applets

Note that certain applets such as oneslime.net regressed slightly 
because of the new Java2D work.  I'm more concerned that the 
appletviewer and plugin still work, and that they're installed in the 
right places:

gappletviewer should be installed in $prefix/bin
libgcjwebplugin.so should be installed in $prefix/lib*/gcj-4.2.0

Are they build by default? I only used
> --enable-java-awt=gtk
>
> > > I'll try to
> > > finish it tomorrow night. But if someone wants to give it a spin that
> > > would be appreciated. Full patch against svn trunk can be found at:
> > > http://www.klomp.org/mark/classpath/libgcj-classpath-20060719-merge.patch.bz2
> > I'll do a test build on x86-64.  You can also post an RFT to gcj-patches 
> > and Andreas Tobler will try builds on the less-common architectures.
>
> Thanks!
>
> BTW. My notes say:
> - Check gconf dependency stuff.

The patch doesn't apply to GCC trunk because of the new gconf-peer 
directory.  Specifically, the 
classpath/native/jni/gconf-peer/Makefile.in fragment fails to apply, 
since the classpath/native/jni/gconf-peer directory doesn't exist on 
trunk yet.

> - VMPipe/VMChannel are a bit hacked up now.
>
> Any help testing/fixing that appreciated.

I'll look over the gconf dependency configury.

Tom