Re: [gtk-vnc-devel] [PATCH] gtkgl based scaling

[Anthony Liguori <anthony codemonkey ws>]

Daniel P. Berrange wrote:
> On Sat, Feb 02, 2008 at 09:06:41PM +0000, Daniel P. Berrange wrote:
>   
> > On Sat, Feb 02, 2008 at 02:16:42PM -0600, Anthony Liguori wrote:
> >     
> > > I've just committed this patch (it's not in the 0.3.3 release).  Please 
> > > try it out and let me know what you think.
> > >
> > > I've modified gvncviewer so that it has a Full Screen and OpenGL Scaling 
> > > menu option.  The dependency on gtkglext is optional and there's also a 
> > > new optional dependency on libview.  This is only used by gvncviewer 
> > > though so I do not recommend that any distro packages make gtk-vnc 
> > > depend on libview.
> > >       
> > The gtkglext stuff seems to be working prertty nicely - it even works on my
> > r500 ATI card where I don't even have 3d X driver support, or even 2d accel!
> >
> > Sometimes though it does not refresh the screen properly. Asan exmaple, if 
> > i'm using a xterm and say type 'pwd<enter>', i'll see the 'pwd' characters 
> > typed, but not the newline and not the command output - until I type another
> > character or move the mouse. So it seems the screen refreshs may be getting 
> > delayed somehow.
> >     
>
> This patch fixes it for me...
>
> diff -r 2d87caab4637 src/vncdisplay.c
> --- a/src/vncdisplay.c	Sat Feb 02 14:24:08 2008 -0600
> +++ b/src/vncdisplay.c	Mon Feb 04 19:55:57 2008 -0500
> @@ -144,6 +144,12 @@ static gboolean expose_event(GtkWidget *
>  	int x, y, w, h;
>  	GdkRectangle drawn;
>  	GdkRegion *clear, *copy;
> +
> +	GVNC_DEBUG("Expose %dx%d @ %d,%d\n",
> +		   expose->area.x,
> +		   expose->area.y,
> +		   expose->area.width,
> +		   expose->area.height);
>  
>  #if WITH_GTKGLEXT
>  	if (priv->image == NULL && priv->gl_tex_data == NULL)
> @@ -207,6 +213,11 @@ static gboolean expose_event(GtkWidget *
>  		glTexCoord2f(rx,ry);  glVertex3f(wx+ww, wy, 0);
>  		glEnd();
>  		glDisable(GL_TEXTURE_2D);
> +		if (gdk_gl_drawable_is_double_buffered(priv->gl_drawable)) {
> +			gdk_gl_drawable_swap_buffers(priv->gl_drawable);
> +		} else {
> +			glFlush ();
> +		}
>  		gdk_gl_drawable_gl_end(priv->gl_drawable);
>  	} else
>  #endif
>
>
> My tests only hit the glFlush() codepath - dunno if the double buffered
> path is needed - this code snippet is from the gtkglext examples.
>   

How is performance?  I don't have access to the machine that has the 
problem ATM so I can't test myself.  I added a swap_buffers() call 
myself but it ended up killing performance.

Regards,

Anthony Liguori

> Dan.
>