X11 / broken input focus, need so. to reproduce

[Hugo Pereira Da Costa]

On Thursday, October 14, 2010 01:32:07 pm Thomas Lübking wrote:
> The mail was originally sent to the kwin mailing list, but since i got no
> reply i'll try here ;-)
> 
> I just want some confirmation that this is not a local problem.

Same problem here.
Setup is:

KDE:  4.5.71 (4.6 >= 20101005)
(some snapshot from mandriva cooker)

Qt: 4.7.0

XOrg:
Not 100% sure of the versionning here.
kinfocenter says:
Vender Release Number: 10,707,000
Version number 11.0



> If you can spare the time and have KDE installed, please compile the
> testcase, run it and try to activate the window by clicking it (you local
> focus policy is irrelevant! - the only WM that seems to "work" here is
> IceWM...)
> 
> Then pleast reply on success and your XOrg/Qt/KDE version (though KDE
> should actually not matter)
> 
> Thanks alot,
> Thomas
> 
> 
> //compile:
> gcc -o focustest focustest.cpp -I /usr/include \
> -I/usr/include/QtGui -lQtCore -lkdeui
> 
> ------- Original mail with some details -------------------
> 
> This is maybe not really a kwin issue, but might be of interest.
> 
> Since "some" recent update (of Xorg, Qt?) i'm no longer able to set the
> input focus to windows after
> 
> XWMHints *hints = XGetWMHints ( QX11Info::display(), window()->winId() );
> hints->input = false;
> XSetWMHints ( QX11Info::display(), window()->winId(), hints );
> 
> ... and i mean: "not at all".
> Neither using QWidget::activateWindow() (used to work) nor
> KWindowSystem::activateWindow() nor even KWindowSystem::forceActiveWindow()
> I also cannot even override the input focus from the rules anymore. And the
> focus policy (tried none) has no impact either.
> 
> I first thought this might be a recent bug in kwin, but it actually works
> with neither compiz nor metacity nor openbox.
> 
> It /does/ however work with [drumroll] IceWM... =\
> 
> Is anyone aware of recent changes in XWMHints or the clientmessage system?
> Is it a general issue at all or only a localdistro one? (ie: can you
> reproduce this, testcase attached)
> And why could IceWM keep working? (it does not ignore the hint, the
> inactive part of the client does not cause an activation)
> 
> Testcase attached, Workspace::allowClientActivation() figures "Activation:
> Belongs to active application", what /should/ result in a permittance ...
> 
> 
> Cheers,
> Thomas