Re: Threaded animation with cairo_gl backend over a GtkWidget

Hi!
looks like it doesn't work. I've added XInitThreads() and it still
crashing. Also added XLockDisplay and XUnlockDisplay with the same result.
Any idea?
Thanks!


int main (int argc, char *argv[])

{

    if(!XInitThreads())

         return 0;


    gdk_threads_init();

    gdk_threads_enter();

    gtk_init(&argc, &argv);


    GtkWidget *window = gtk_window_new(GTK_WINDOW_TOPLEVEL);

    g_signal_connect(window, "destroy", G_CALLBACK(gtk_main_quit), NULL);

    gtk_window_set_title(GTK_WINDOW(window), "cairo_gl");

    gtk_window_set_default_size(GTK_WINDOW(window), 400, 300);

    gtk_widget_set_app_paintable(window, TRUE);

    gtk_widget_set_double_buffered(window, FALSE);


    gtk_widget_show_all(window);

     window_surface=create_source_surface_for_widget(window);

     (void)g_timeout_add(33, (GSourceFunc)timer_exe, window);

     //do_draw(NULL);


    gtk_main();


    gdk_threads_leave();

    return 0;

}


void *do_draw(void *ptr)

{

     rendering=TRUE;

     cairo_t *cr = cairo_create(window_surface);

  if(cairo_status(cr))

return NULL;

   Display *dpy =cairo_glx_device_get_display((cairo_surface_get_device(
window_surface)));

     XLockDisplay(dpy);

    //do some time-consuming drawing here ...

    // ...

    // ...

  cairo_destroy(cr);


    cairo_gl_surface_swapbuffers (window_surface);

   XUnlockDisplay(dpy);

   rendering=FALSE;


    return NULL;

}




2013/1/11 Carlos López González <[email protected]>

> Thanks Henry, I'll give it a try.
>
>
> 2013/1/11 Henry (Yu) Song - SISA <[email protected]>
>
> Hi,
>>
>> Please add XInitThreads() in your main.  For multithreaded application
>> using gl_surface is not a good idea for the moment unless you have
>> intel/radeon/nouveau drivers, otherwise, you will have very sloooooow
>> performance.
>>
>> We will resolve that soon in cairo for gl backend, I hope.
>>
>> Henry
>> ________________________________________
>> From: [email protected][cairo-bounces+henry.song=
>> [email protected]] on behalf of Carlos López González [
>> [email protected]]
>> Sent: Thursday, January 10, 2013 11:26 AM
>> To: [email protected]
>> Subject: [cairo] Threaded animation with cairo_gl backend over a GtkWidget
>>
>> Hi!
>> I've made a modified version [1] of the example of threaded animation
>> using Cairo [2].
>>
>> First I've used the method to create a cairo_gl_surface from the
>> test/gl_source_surface.c file adapted to create the surface for a given
>> widget instead for a defined size.
>>
>> Also, for simplicity, I removed all the handled signals except the
>> destroy one.
>>
>> To handle the drawing for the window widget using the cairo_gl backend I
>> use the create_gl_surface_for_window function passing it the proper widget
>> information.
>>
>> So far, for a single drawing execution it works as expected:
>> 1) In main() create the window_surface
>> 2) The do_draw function creates a context based on window_surface
>> 3) Once done the drawings with Cairo then the buffers are swapped and the
>> content is shown on screen.
>>
>> Notice that the window_surface is never destroyed explicitly because it
>> is owned by the window widget and so it will be destroyed when the window
>> is destroyed by gtk.
>>
>> The problem comes when I enable the multithreading support by
>> uncommenting the line:
>>
>>
>> //(void)g_timeout_add(33, (GSourceFunc)timer_exe, window);
>>
>> It triggers the timer_exe function 33 times per second and then it
>> creates threads that will call do_draw each time.
>>
>> Once enabled it crashes at the first painting function from Cairo.
>>
>> I believe that the problem is that the window_surface belongs to the
>> window widget when I created it and then I can't modify it outside the main
>> gtk thread.
>>
>> How can I solve this? If I enclose the Cairo painting operations inside a
>> gdk_threads_enter and
>> gdk_threads_leave I would add a lot of time consuming operations to the
>> main gtk thread, which is precisely what I want to avoid using
>> multithreading.
>>
>> Any help? I would like to know at last, if my assumptions of the
>> window_surface ownership are true or not.
>>
>> Thanks!
>>
>> [1]
>> https://github.com/genete/my-test-area/blob/direct_draw/src/cairo_sample.c
>> [2] http://cairographics.org/threaded_animation_with_cairo/
>>
>> --
>> Carlos
>> http://synfig.org
>>
>
>
>
> --
> Carlos
> http://synfig.org
>



-- 
Carlos
http://synfig.org

-- 
cairo mailing list
[email protected]
http://lists.cairographics.org/mailman/listinfo/cairo