Independent clone mode

[Pekka Paalanen]

On Wed, 31 Jul 2019 09:48:07 +1200
Barry Song <21cnbao at gmail.com> wrote:

> Pekka Paalanen <ppaalanen at gmail.com> 于2019年7月27日周六 下午8:20写道：
> >
> > On Fri, 26 Jul 2019 08:58:42 +1200
> > Barry Song <21cnbao at gmail.com> wrote:
> >  
> > > Pekka Paalanen <ppaalanen at gmail.com> 于2019年7月26日周五 上午3:30写道：  
> > > >
> > > > On Thu, 25 Jul 2019 13:00:55 +1200
> > > > Barry Song <21cnbao at gmail.com> wrote:
> > > >  
> > > > > Pekka Paalanen <ppaalanen at gmail.com> 于2019年7月24日周三 下午8:10写道：  
> > > > > >
> > > > > > On Wed, 24 Jul 2019 13:22:43 +0800
> > > > > > zou lan <nancy.lan.zou at gmail.com> wrote:
> > > > > >  
> > > > > > > Hi pekka
> > > > > > >
> > > > > > > I see the clone mode is supported from this commit message.
> > > > > > > https://patchwork.freedesktop.org/patch/227970/
> > > > > > >
> > > > > > > I also see the message  "Independent CRTC clone mode cannot be supported
> > > > > > > until output layout logic is moved from libweston into the frontend and
> > > > > > > libweston's damage tracking issues stemming from overlapping outputs are
> > > > > > > solved".
> > > > > > >
> > > > > > > I want to ask is independent CRTC clone mode already supported in Weston  6
> > > > > > > or 7? If not,
> > > > > > > is there a plan to support it?  
> > > > > >
> > > > > > Hi,
> > > > > >
> > > > > > no, that commit message is still accurate. I don't know of any
> > > > > > plans to work on independent CRTC clone mode.  
> > > > >
> > > > > Pekka,
> > > > > any possible workaround do you know to support independent crtc clone
> > > > > within weston? right now it is very important product feature.
> > > > > for example, force weston to identify only one screen, and sync copy
> > > > > the screen to 2nd screen and call the drm api of 2nd screen?  
> > > >
> > > > Hi,
> > > >
> > > > you could hack it up to do that, but it would be in no way an
> > > > upstreamable solution. It would also hurt display timings because
> > > > every repaint would have to wait for both monitors to refresh. Or,
> > > > all but one monitor would tear.
> > > >
> > > > A simpler hack would be to create overlapping outputs and
> > > > force them repaint fully on every refresh instead of repainting
> > > > only the damage. That's not upstreamable either, but the damage
> > > > could be hacked with probably a one-liner.  
> > >
> > > Hi Pekka,
> > > Thanks. even though i have seen many times that damage tracking of
> > > overlapping outputs is the main cause stopping weston from supporting
> > > full clone mode, i haven't fully understood what is overlapping
> > > outputs in weston and found the code related with overlapping output,
> > > would you like to point out some code path so that i can begin to do
> > > some work on it? pls forgive me if you think the question is just
> > > stupid.  
> >
> > Hi,
> >  
> 
> 
> i made a very great progress to hack weston to support clone mode in
> the direction you pointed out.
> 
> it works very well without any tearing.

Hi,

sorry, I did not mean tearing. I meant areas with outdated content.

I guess the best way to make that clearly visible is to have the
outputs overlap by 50%, not 100%, and then continuously move a window
such that it will cross all four areas:
- output A: exclusive area
- output A: shared with output B area
- output B: shared with output A area
- output B: exclusive area

Then observe inconsistency in the image between the exclusive and the
shared area on each output separately. If you see the window decimate
on either output, that's bad.

Mind that the effect is semi-random, because it depends on the timings
of the outputs and the client and desktop updates.


> > what I suggested is not about code for overlapping outputs at all.
> > I'm simply suggesting to hack e.g. weston_output_repaint() or
> > something to ignore the accumulated damage and use full damage
> > instead. I can't tell you exactly what and where without doing it
> > myself first, which I cannot do at the moment.  
> 
> hacked by:
> 
> @@ -2406,13 +2406,20 @@ weston_output_repaint(struct weston_output
> *output, void *repaint_data)
>                 }
>         }
> 
> -       compositor_accumulate_damage(ec);
> +       if (!output->external) {
> +               compositor_accumulate_damage(ec);
> 
> -       pixman_region32_init(&output_damage);
> -       pixman_region32_intersect(&output_damage,
> -                                 &ec->primary_plane.damage, &output->region);
> -       pixman_region32_subtract(&output_damage,
> -                                &output_damage, &ec->primary_plane.clip);
> +               pixman_region32_init(&output_damage);
> +               pixman_region32_intersect(&output_damage,
> +                               &ec->primary_plane.damage, &output->region);
> +               pixman_region32_subtract(&output_damage,
> +                               &output_damage, &ec->primary_plane.clip);
> +       }  else {
> +               /** From Pekka Paalanen: the damage could be hacked
> with probably
> +                * a one-liner, ignore the accumulated damage and use
> full damage
> +                */
> +               output_damage = output->region;
> +       }

That is not enough, unless you also stopped clearing the damage in the
external case.

That assignment is also wrong: assigning regions like that is prone to
memory corruption.

> since so many people are asking for clone mode in weston. would this
> hacking version be accepted as an initial patch to support independent
> clone?

I would not take it, sorry.


Thanks,
pq
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 833 bytes
Desc: OpenPGP digital signature
URL: <https://lists.freedesktop.org/archives/wayland-devel/attachments/20190802/6c95435f/attachment.sig>