[21258] in Athena Bugs
Re: linux 9.1.18: mozilla
daemon@ATHENA.MIT.EDU (John Carr)
Sun Dec 29 15:43:00 2002
Message-Id: <200212292042.PAA16451@psi-phi.mit.edu>
To: Greg Hudson <ghudson@MIT.EDU>
cc: bugs@MIT.EDU
In-Reply-To: Your message of "21 Dec 2002 18:42:36 EST."
<1040514156.10241.156.camel@error-messages.mit.edu>
Date: Sun, 29 Dec 2002 15:42:58 -0500
From: John Carr <jfc@MIT.EDU>
> On Sat, 2002-12-21 at 11:53, John Carr wrote:
> > The setting of input focus is supposed to be invisble to clients.
> > The only way for a client to tell which policy is in use is to
> > call XGetInputFocus.
>
> Mozilla does not appear to use XGetInputFocus except in one obscure
> place having to do with popups. So by your reasoning, mozilla should
> not be able to behave differently with PointerRoot.
You are correct, XGetInputFocus is not the problem. It turns out
there is a second difference between active and passive focus. Focus
events are sent by the X server when the window manager forces focus
to a window but not when a window acquired focus as a result of mouse
movement.
gtk attempts to handle windowmanagerless input by responding to both
focus events and enter/leave events, but fails to request that
enter/leave events be delivered. Not having noticed that the pointer
entered the window, gtk rejects keyboard input.
This change to gtk/gdk appears to fix the bug -- I can type into the
mozilla window. I haven't done any significant testing for adverse
effects. (It's still a horrid program, but I'm now able to experience
the intentional flaws instead of the accidental ones.)
*** /mit/source-9.1/third/gtk/gdk/gdkwindow.c Wed Mar 14 12:11:00 2001
--- gdkwindow.c Sun Dec 29 15:34:28 2002
***************
*** 58,64 ****
KeyReleaseMask,
EnterWindowMask,
LeaveWindowMask,
! FocusChangeMask,
StructureNotifyMask,
PropertyChangeMask,
VisibilityChangeMask,
--- 58,64 ----
KeyReleaseMask,
EnterWindowMask,
LeaveWindowMask,
! FocusChangeMask|EnterWindowMask|LeaveWindowMask,
StructureNotifyMask,
PropertyChangeMask,
VisibilityChangeMask,
