On Mon, 6 Sep 2021 11:05:42 GMT, Jose Pereda <jper...@openjdk.org> wrote:
>> I don't doubt that the area should be cleared. What I was questioning is
>> whether this was the best place to do it? I'd be OK with it if you document
>> it.
>
> The two options I see so far to clear the area are the one I've committed
> (change in `WCGraphicsPrismContext::setClip`) and the one I've mentioned
> above (`WCPageBackBufferImpl::copyArea`).
> After some more testing, I see that the latter doesn't work when there is
> full transparency, so I will discard it.
>
> Back to my original commit, it could be moved up to `WebPage::paint2GC`,
> before the call to `gc.setClip(rq.getClip());`, and that would also remove
> the need of passing down the opacity level to `WCGraphicsContext`. I also see
> now that the `clearRect()` call is required for full transparency only.
>
> This would be the change:
>
> WCRectangle clip = rq.getClip();
> if (clip != null) {
> if (isBackgroundTransparent()) {
> // As backbuffer is enabled, new clips are drawn over
> the old rendered frames
> // regardless the alpha channel. While that works
> fine for alpha > 0,
> // for alpha == 0 we need to clear the old frame or
> it will still be visible.
>
> gc.clearRect((int) clip.getX(), (int) clip.getY(),
> (int) clip.getWidth(), (int) clip.getHeight());
> }
> gc.setClip(clip);
> }
This seems like the best approach to me in that it moves the clear to a drawing
method rather than as a side effect of setting the clip.
-------------
PR: https://git.openjdk.java.net/jfx/pull/563
