Hi; thanks for the helpful explanations.
I tried doing that change and it stayed the same. I also tried doing a
sleep(2) before sending the WM_TAKE_FOCUS (with the +2s timestamp), but without
success.
What did work is this:
static void clientwin_do_set_focus(WClientWin *cwin, bool warp)
{
if(cwin->flags&CLIENTWIN_P_WM_TAKE_FOCUS){
Time stmp=ioncore_get_timestamp();
send_clientmsg(cwin->win, ioncore_g.atom_wm_take_focus, stmp);
} else {
region_finalise_focusing((WRegion*)cwin, cwin->win, warp);
}
XSync(ioncore_g.dpy, 0);
}
I just added the „else“ keyword so that region_finalise_focusing is not
called if the WM_TAKE_FOCUS is sent. I don't know the meaning of this, but only
that it solves all problems with Netbeans and my example Swing application.
Focus is set to the last active widget.
This, however, makes Konqueror 4 stop receiving focus. This makes me think:
- Konqueror has flag CLIENTWIN_P_WM_TAKE_FOCUS and apparently needs both the
WM_TAKE_FOCUS call and the region_finalise_focusing
- Netbeans/Swing also has flag CLIENTWIN_P_WM_TAKE_FOCUS, but if you run the
region_finalise_focusing code on it, it loses focus
Maybe region_finalise_focusing should sometimes be doing nothing in windows
like Netbeans.
Or maybe Swing should be correctly handling the XSetInputFocus(ioncore_g.dpy,
win, RevertToParent, CurrentTime) in region_finalise_focusing.
It is exactly at that XSetInputFocus where the text field at the Swing
application loses the keyboard focus (I tested it inserting sleeps between
calls).
I filed this bug to Sun but I don't know whether they think the bug is in
JDK. The bug report is at http://bugs.sun.com/view_bug.do?bug_id=6798064
Greetings,
Daniel
