On 21-01-2026 19:28, Xaver Hugl wrote:
Oh, I just noticed: It would help if we had a CAP for this, so the
compositor knows if the new API is supported before doing atomic tests
with it.
No new UAPI is added for this and would be enabled by default with
atomic ioctl.
The reserved field in the struct drm_mode_atomic would be used to return
the error code. If this value is zero with a atomic_ioctl failure then
eventually should be a older kernel where this feature is not added. For
every failure on atomic_ioctl an error_code would be sent by the kernel.
Let me know if you still think a UAPI is required to get to know the
capability of this feature.
Thanks and Regards,
Arun R Murthy
--------------------
