Re: [cp-patches] FYI: Some quick fixes for accessibility in Swing

[Mark Wielaard <mark at klomp dot org>]

Hi,

On Wed, 2005-09-07 at 14:49 +0200, Roman Kennke wrote:
> Am Mittwoch, den 07.09.2005, 07:44 -0400 schrieb Thomas Fitzsimmons:
> > On Tue, 2005-09-06 at 22:41 +0200, Roman Kennke wrote:
> > > I re-enabled some accessibility things in javax.swing that have been
> > > disabled due to a compiler bug in gcj < 4. I think now that the 0.18
> > > release is out it would be a good time to get rid of this
> > > gcj<4-bug-support things. This will also remove a lot of clutter in the
> > > JAPI pages and maybe brings some points in the score :-)
> > 
> > This still breaks with gcj HEAD:
> > 
> > /home/fitzsim/sources/gcc/libjava/classpath/javax/swing/Box.java:67:
> > error: Nested class java.awt.Container$AccessibleAWTContainer is
> > protected; cannot be accessed from here.
> >   protected class AccessibleBox extends Container.AccessibleAWTContainer
> >                                            ^
> > /home/fitzsim/sources/gcc/libjava/classpath/javax/swing/Box.java:92:
> > error: Nested class java.awt.Component$AccessibleAWTComponent is
> > protected; cannot be accessed from here.
> >       extends Component.AccessibleAWTComponent
> > 
> > I assume this is a known GCC bug?
> 
> I double checked this. It definitly does compile with the gcj-4.0 that
> ships with ubuntu (a pre-something release, but anyway. I assume there
> have been no regressions between 4.0-preX and 4.0-final). So this looks
> like a regression in gcj4.1/HEAD.

Disturbing. It also breaks (in the same way as reported above) with
'gcj (GCC) 4.0.2 20050908 (prerelease)' and 'gcj (GCC) 4.0.2 20050821
(prerelease) (Debian 4.0.1-6)' so this seems to be a regression from
4.0-preX to 4.0.x.

I am inclined to just make the AccessibleAWTContainer and
AccessibleAWTComponent public. while a little wrong at least it makes us
able to compile again.

Tom Tromey said he thought this was PR8544, but that seems to be a
(same-)package visibility bug that can be worked around by fully
qualifying the inner interface name. But this is an access from another
package. I haven't found a better PR though. There are many mentioned in
the inner-class meta-bug though, so maybe it is already filed:
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18131

Cheers,

Mark

Attachment: signature.asc
Description: This is a digitally signed message part