[PATCH v15 03/34] compositor-drm: Move repaint state application to flush

[Daniel Stone]

Hi,

On 6 February 2018 at 12:25, Pekka Paalanen <ppaalanen at gmail.com> wrote:
> I found some little details to fix, but other than those, it looks good.

Thanks! For the record, I've fixed all these before pushing.

> One thing I noticed in the legacy KMS path is that we don't tend to
> program DPMS state when turning the CRTC off. Does the kernel do the
> right thing, or should we always do DPMS-off when doing CRTC-off?
>
> I'm more interested in the theory there, I don't mind if we leave the
> monitor saying "no signal". Atomic will get that right anyway, I
> presume, and monitors will probably turn off by themselves.

Hmm. I actually need to re-check that with more drivers having moved
to atomic (even if we use the legacy API, it still has an impact), but
my recollection is that they _should_ be one and the same. Of course,
it might make sense to just program the property anyway.

I've jotted that down to take a closer look at (including the DPMS
property enum/type confusion) next week. The other thing I noticed
which was broken (not at all new, but slightly more failure-prone with
very recent kernels) is MST which results in connectors appearing and
disappearing. I think it's worth holding off until we land the clone
mode foundation series to try to tackle that though.

Cheers,
Daniel