Hi Jeff:  Any ideas on this one?  The way I can reproduce it is by running any Java application (I ran SwingSet2) -- with JDK 1.6, tab will move you forward, but shift+tab will not move you backwards.  With JDK 1.5, things work as expected.  Here's the specific JDK's I'm using:

bash-3.2$ uname -a; /usr/jdk/jdk1.6.0_01/bin/java -version; /usr/jdk/jdk1.5.0_12/bin/java -version
SunOS solaris-devx 5.11 snv_76 i86pc i386 i86pc Solaris
java version "1.6.0_01"
Java(TM) SE Runtime Environment (build 1.6.0_01-b06)
Java HotSpot(TM) Client VM (build 1.6.0_01-b06, mixed mode, sharing)
java version "1.5.0_12"
Java(TM) 2 Runtime Environment, Standard Edition (build 1.5.0_12-b04)
Java HotSpot(TM) Client VM (build 1.5.0_12-b04, mixed mode, sharing)

I'm also using the latest at-spi and JABG from SVN trunk.  This also occurs regardless of whether Orca is running or not.

When you press Tab key, java bridge receives three key events:

KeyEvent.KEY_PRESSED
KeyEvent.KEY_TYPED
KeyEvent.KEY_RELEASED

When you press Shift-Tab key, java bridge only receives two key events:

KeyEvent.KEY_PRESSED
KeyEvent.KEY_RELEASED

Under this case, java bridge will not forward KeyEvent.KEY_PRESSED event to applications, only KEY_RELEASED is sent out. This is the reason that java applications can't respect to Shift-Tab

Created attachment 98708 [details] [review]
Forward the pending event before fowarding the next event

Willie and Lynn, could you have a test on a package with the patch? If there is no regression, I can commit it.

(In reply to comment #4)
> Willie and Lynn, could you have a test on a package with the patch? If there is
> no regression, I can commit it.

This gives me Shift+Tab back.  Thanks!