[soreau]

This fixes broken rendering for blurred and wobbly surfaces on outputs transformed 90 and 270.

Fixes #3161.

[EoceneMiacid]

Interesting. Disabling the 'blur' plugin does indeed fix the problem in the sense that apps now maximize correctly on it, so the rotated screen is at least usable now.

But it would appear there's another problem - here's a video that showcases it > https://imgur.com/NJz8UIH

This is when dragging an unmaximized window on it, only the upper half is used.

plugins=ipc ipc-rules autostart alpha animate command move switcher resize place zoom fisheye place wobbly window-rules wf-info xdg-activation wm-actions csd-decorator

[soreau]

Interesting. Disabling the 'blur' plugin does indeed fix the problem in the sense that apps now maximize correctly on it, so the rotated screen is at least usable now.

But it would appear there's another problem - here's a video that showcases it > https://imgur.com/NJz8UIH

This is when dragging an unmaximized window on it, only the upper half is used.

I'm pretty sure the bug in your video is precisely the bug that this fixes. I would suggest either double checking that blur is disabled or to install the one-liner patch in this pull request.

[soreau]

Ah, I didn't have wobbly enabled. Seems like we'll need this for wobbly too:

diff --git a/plugins/wobbly/wobbly.cpp b/plugins/wobbly/wobbly.cpp
index ea0da710..e314def6 100644
--- a/plugins/wobbly/wobbly.cpp
+++ b/plugins/wobbly/wobbly.cpp
@@ -1004,7 +1004,7 @@ class wobbly_render_instance_t :
             wf::gles::for_each_scissor_rect(data.target, data.damage, [&]
             {
                 wobbly_graphics::render_triangles(self->wobbly_program, tex,
-                    wf::gles::render_target_aligned_orthographic_projection(data.target),
+                    wf::gles::render_target_orthographic_projection(data.target),
                     vert.data(), uv.data(), count_triangles);
             });
         });

[EoceneMiacid]

Yes - disabling 'wobbly' fixes that too - nice catch. It's almost perfect now.

Only one problem remains, which is that the vertical size doesn't align properly between the rotated and the non-rotated screen > https://imgur.com/4aD42YD

The left screen is 1920 pixels high sitting next to a screen 2160 pixels high, yet it will only allow me to move the mouse cursor from the left to the right screen in the upper half.

[soreau]

Only one problem remains, which is that the vertical size doesn't align properly between the rotated and the non-rotated screen > https://imgur.com/4aD42YD

The left screen is 1920 pixels high sitting next to a screen 2160 pixels high, yet it will only allow me to move the mouse cursor from the left to the right screen in the upper half.

This is a different matter entirely for which you should file a separate issue.

[EoceneMiacid]

Ok will do

[ammen99]

This is the wrong fix. I have made these changes particularly so that rendering remains crisp with fractional scale - moving back to the not-aligned versions will break that.

Somehow, we need to ensure that alignment works the same across rotated outputs. Maybe we'll have to invert the order of scaling and rotating in this path, or some other trick to ensure alignment works as it should.

[soreau]

This is the wrong fix. I have made these changes particularly so that rendering remains crisp with fractional scale - moving back to the not-aligned versions will break that.

Somehow, we need to ensure that alignment works the same across rotated outputs. Maybe we'll have to invert the order of scaling and rotating in this path, or some other trick to ensure alignment works as it should.

Thank you for your review. I think I found the real culprit. Please check the latest diff.

[ammen99]

LGTM now, thanks!

[ammen99]

Could you please rebase so that the CI can run and we make sure style is ok?