Daniel Laird wrote:
>> 1) Find out why the background is cleared at all, because
>> layers != PRIMARY should have no background color set,
>> or maybe it's black. The background mode for these layers
>> should be DLBM_DONTCARE to solve the issue.
>>
>> 2) Not create a window stack when the layer has a certain
>> capability, maybe add DLCAPS_NOWINDOWSTACK.
>>
>> 3) Allocate a dummy line of pixels in your surface pool and
>> keep the pitch at 0 :-)
>>
> Option 3 is probably the best but not the cleanest. I had a quick go with
> option 2.
Option 3 is at least the one with the least impact or risks.
> I modified dfb_layer_context_init in layer_context.c.
> I added
> /* Create the window stack if the layer supports it. */
> if (!D_FLAGS_IS_SET( shared->description.caps, DLCAPS_NOWINDOWSTACK ))
> {
> context->stack = dfb_windowstack_create( context );
> if (!context->stack) {
> dfb_layer_context_unref( context );
> return D_OOSHM();
> }
>
> /* Tell the window stack about its size. */
> dfb_windowstack_resize( context->stack,
> context->config.width,
> context->config.height );
> }
> This seemed to do the job rather nicely. Can you see any drawbacks? know of
> any hidden issues I might encounter.
The issues could be wherever context->stack is assumed to be non-NULL,
which should cause failing assertions, but some sites might just cause
a segmentation fault.
So option 1 would be quite clean and somewhat safe, but at any time a
user might request to create a window or render to the layer surface
directly...
I think I'd go for Option 1+3 which avoids filling the dummy line
multiple (height) times, but still has a fallback in case there's
another access.
Denis
_______________________________________________
directfb-dev mailing list
directfb-dev@directfb.org
http://mail.directfb.org/cgi-bin/mailman/listinfo/directfb-dev
