[xine-devel] bug in alphablending (was: [dvd-devel] Problems with Birdcage DVDs.)

[Michael R. <mr...@us...>]

taking this over to xine-devel, because this is clearly a bug in xine

Hi,

> >>Seems like it's still the same. I tested only xv and xshm with 16
> >> Bit as they both had the very same symptoms. I also retested the
> >> Herbert Gr=F6nemeyer DVD and I aslo stumbled across two more DVD
> >> which show the same behaviour: Sixth Sense RC2 german and The
> >> Terminator Special Ediition RC2 UK.
> >
> >I have the Sixth Sense dvd here, maybe I can help: Could you explain
> > the problem with this disc again?
>
> The problem with the discs is that the highlighting is wrong. i.e. if
> I move the mouse to item 2 (top -> bottom) item 1 is highlighted.
> Being over item 1 nothing is highlighted. This broke somewhere
> between 0.9.12 and 0.9.13

Sorry that this took so long, but I finally found time to look into=20
this. The current alphablend routines seem to be wrong (have no idea=20
when this broke).
Problem is the clipping coordinates in the overlay structure: The code=20
seems to be uncertain about the coordinate system these coordinates are=20
referring to. Example? Ok, here we go:

This is blend_rgb24, line 547:
  /* avoid wraping overlay if drawing to small image */
  if( (img_overl->x + img_overl->clip_right) < dst_width )
    clip_right =3D img_overl->clip_right;
  else
    clip_right =3D dst_width - 1 - img_overl->x;
This only makes sense, if the clipping coordinates are relative to the=20
left upper corner of the OVERLAY.

But this is blend_rgb24, line 580:
        } else if( clip_right < x + rlelen ) {
          if( clip_right > x ) {
This suggests that the clipping coordinates are relative to the left=20
upper corner of the FRAME (since x iterates over frame width).

So I think we should decide now, which behaviour (frame vs. overlay=20
coordinates) we want. Are there any performance issues with one of=20
them? Libspudec currently assumes overlay coordinates. (This is why the=20
user sees the reported bug.)

Michael

--=20
printk("ufs_read_super: fucking Sun blows me\n");
	2.0.38 /usr/src/linux/fs/ufs/ufs_super.c