On Fri, Mar 01, 2024 at 01:12:02PM +0100, Maxime Ripard wrote: > On Fri, Mar 01, 2024 at 12:29:41PM +0100, Sebastian Wick wrote: > > On Fri, Mar 01, 2024 at 11:30:56AM +0100, Maxime Ripard wrote: > > > On Thu, Feb 29, 2024 at 08:47:26PM +0100, Sebastian Wick wrote: > > > > > @@ -1708,6 +1731,39 @@ EXPORT_SYMBOL(drm_connector_attach_dp_subconnector_property); > > > > > /** > > > > > * DOC: HDMI connector properties > > > > > * > > > > > + * Broadcast RGB (HDMI specific) > > > > > + * Indicates the Quantization Range (Full vs Limited) used. The color > > > > > + * processing pipeline will be adjusted to match the value of the > > > > > + * property, and the Infoframes will be generated and sent accordingly. > > > > > + * > > > > > + * This property is only relevant if the HDMI output format is RGB. If > > > > > + * it's one of the YCbCr variant, it will be ignored and the output will > > > > > + * use a limited quantization range. > > > > > > > > Uh, maybe just say that the quantization range is selected automatically > > > > in case a YCbCr output format is in use. I'm not sure every YCbCr > > > > variant requires limited and even if it does, new formats could change > > > > this. > > > > > > I documented what i915 is doing: > > > https://elixir.bootlin.com/linux/latest/source/drivers/gpu/drm/i915/display/intel_hdmi.c#L2143 > > > > Sure, this is one valid strategy for the automatic behavior of YCbCr. > > Drivers could also always send an InfoFrame to ensure full range where > > possible. The point here is that this property shall not affect YCbCr > > output formats! > > > > Maybe it's even better to say "driver specific" instead of "automatic". > > Honestly, I'm not sure what you want from me here. Ville and you > insisted on the previous version to document what i915 is doing and to > follow whatever the behaviour was, and that we shouldn't spend time > improving the property. Fine, I did that. > > But now, you want me to ... improve the property? The property has a clear scope: quantization range for RGB output formats. What Intel does with things that are not in scope of the property is irrelevant. This isn't improving the property either but documenting the scope of the property. Sorry if this seems arbitrary but these details are so important to get right because the interactions between all the different things is already majorly broken. > Maxime >
