Simon Ser

It's pretty barebones atm.

See https://tools.ietf.org/html/rfc4155

Right now we re-render each time we receive a sequence event. This means we render too often, especially if there are multiple active CRTCs. Instead, render at a fixed rate...

See [1]. [1]: https://github.com/swaywm/wlroots/pull/1985

- [x] Dumb buffer allocator - [ ] Pixman renderer - [ ] Check for `DRM_CAP_DUMB_BUFFER` - [ ] Use `DRM_CAP_DUMB_PREFERRED_DEPTH` - [ ] Use `DRM_CAP_DUMB_PREFER_SHADOW`

This is necessary for damage tracking.

Consider adding get_gem_handle to buffer interface

1

Without this, we export and import our own buffers. We also require DMA-BUF support in the driver. Not sure it's that of a big deal.

Design a libliftoff-friendly backend API

1

Right now the compositor just accesses libliftoff's state directly. That's definitely not a great interface. We should design a backend API that works well with libliftoff. Related to https://github.com/swaywm/wlroots/issues/1826.