netbeans broken on updated Fedora 9

[Per Bothner]

NetBeans (see netbeans.org) version 6.1 (but other versions also fail)
recently stopped working in any useful way after recently updating
Fedora 9.  It works in Fedora 8, when using the identical
copy of NetBeans and JDK.  This suggests either a Gnome/X11/??? bug,
or that JDK is depending on unspecified Gnome/X11/??? behavior.
I'm only seeing this in NetBeans; OTOH I don't use many other
graphical Java applications.

The bug is that redisplay of source editor windows is bad.
I make a change, scroll, or just click and it will most of the
time repaint wrong, showing both the re-painted text and a copy
of some other part of the screen, on top of each other.  This
makes NetBeans on F9 practically unusable

When I use OpenJDK (as in /home/bothner/Java/netbeans-6.1/bin/netbeans 
--jdkhome /usr/lib/jvm/java-1.6.0-openjdk) I got no error message.

If I use jdk1.6.0_05 I do get some error messages in the terminal
window.  On start-up I get:

Locking assertion failure.  Backtrace:
#0 /usr/lib/libxcb-xlib.so.0 [0xc92767]
#1 /usr/lib/libxcb-xlib.so.0(xcb_xlib_unlock+0x31) [0xc92831]
#2 /usr/lib/libX11.so.6(_XReply+0x244) [0xbb8f64]
#3 /opt/jdk1.6.0_05/jre/lib/i386/xawt/libmawt.so [0xb4b508ce]
#4 /opt/jdk1.6.0_05/jre/lib/i386/xawt/libmawt.so [0xb4b2d067]
#5 /opt/jdk1.6.0_05/jre/lib/i386/xawt/libmawt.so [0xb4b2d318]
#6 
/opt/jdk1.6.0_05/jre/lib/i386/xawt/libmawt.so(Java_sun_awt_X11GraphicsEnvironment_initDisplay+0x2f) 
[0xb4b2d61f]
...

and on each re-display I get one or more of:

(<unknown>:3174): Gtk-CRITICAL **: gtk_paint_box: assertion 
`style->depth == gdk_drawable_get_depth (window)' failed

If I use JDK 5 it's even worse: The "status" messages in the
start-up window are all written on top of each other, and when
the main window is up it is blank.

This is on a Dell Inspiron E1405 on an Intel Corporation Mobile 
945GM/GMS, 943/940GML Express Integrated Graphics Controller rev 3.

I saw something similar early in Fedora 9 beta, but the problem
went away, but then it recently returned.

I can report this in BugZilla, but this seems to be a recent
regression, so it seems worthwhile mentioning it here.
-- 
	--Per Bothner
per at bothner.com   http://per.bothner.com/