In D24112#546605, @romangg wrote:

In D24112#546346, @bshah wrote:

@romangg Did you look into last comment? Commit 26cdfd317fc87b024ebcd9c2adefc86cfd59504e was added to recover from the broken state of rotation/bad atomic commit. It seems that this code has regressed (atomic commit related code), we are not reverting back to correct plane (or output size), explicitly setting the correct mode using kscreen-doctor does recover from broken state/hang.

Which comment?

@romangg Did you look into last comment? Commit 26cdfd317fc87b024ebcd9c2adefc86cfd59504e was added to recover from the broken state of rotation/bad atomic commit. It seems that this code has regressed (atomic commit related code), we are not reverting back to correct plane (or output size), explicitly setting the correct mode using kscreen-doctor does recover from broken state/hang.

Commited to 5.17, should I cherry pick the commit to master or merge the 5.17 branch back into master?

Please commit to 5.17 as well.

In D24112#538658, @romangg wrote:

Thanks for looking into this. But I don't think there is a (good) solution possible without some principle changes. For example wayland output device transformation is set in the Drm backend but should be set on AbstractWaylandOutput level.

Overall I want to have the following structure:

Also one of my initial suspicion was that we are not setting transformation wl_output in case commit fails:

Some more debugging with master branch,

However, this also means that the drm backend does bad job at state management (TEST_ONLY commit should have failed).

What I meant by raspberry-pi usecase being weird is, if you have external case for raspberry pi and connect display then it is "internal" in a way. I am not entirely sure what definition of internal we are following here.

Raspberry-PI usecase is slightly weird, but apart from that almost all of the mobile and ARM tablet devices uses MIPI DSI connector. So not really no, you can't make it much more specific.

This bit is mostly unused apart from the orientation sensor enabling, but nevertheless, I want to commit this to Plasma/5.17 branch (to make it easier for me to test rotation patch further on some of target devices)

Looks good to me.

Whoops, missed this and committed a fix already

As mentioned on IRC, +1 to this patch. If this violates a coding style, we can discuss it in the mailing-list/IRC later and either

See fbda1d799cc1b19c3aeb0ac9d445477e871ae85e in attica

Any update on voting process @lydia ?