On Mon, Nov 13, 2023 at 03:23:15AM +0200, Laurent Pinchart wrote:
> On Sun, Nov 12, 2023 at 08:32:33PM +0000, Sakari Ailus wrote:
> > On Sun, Nov 12, 2023 at 03:22:46AM +0200, Laurent Pinchart wrote:
> > > On Fri, Nov 10, 2023 at 09:15:52PM +0000, Sakari Ailus wrote:
> > > > On Fri, Nov 10, 2023 at 05:39:40PM +0200, Laurent Pinchart wrote:
> > > > > On Fri, Nov 10, 2023 at 12:10:49PM +0200, Sakari Ailus wrote:
> > > > > > Compile sub-device state information access functions
> > > > > > v4l2_subdev_state_get_{format,crop,compose} unconditionally as they are
> > > > > > now also used by plain V4L2 drivers.
> > > > >
> > > > > What do you mean by "plain" V4L2 drivers here ?
> > > >
> > > > Those that do not need MC: if you now disable MC in kernel configuration,
> > > > these functions won't be available.
> > >
> > > This covers subdev drivers (such as sensor drivers) too, not just host
> > > drivers ? If so, shouldn't we also drop the dependency on
> > > VIDEO_V4L2_SUBDEV_API from drivers/media/i2c/Kconfig ?
> >
> > Good question.
> >
> > There are still other functions that a number of drivers depend on which
> > are only available when either MC or sub-device API are enabled.
> >
> > Given that almost all receiver drivers today use MC, I don't think there
> > would be much to be gained from enabling building e.g. sensor drivers
> > without sub-device UAPI.
>
> I would like to refocus VIDEO_V4L2_SUBDEV_API to cover the UAPI only.
> Trying to be too clever with the Kconfig symbol doesn't help.
I guess this has been the intention, with the inclusion of other functions
and definitions that are only needed for sub-device UAPI. This poses some
issues for drivers that can work in both cases but there aren't many of
them. I'm fine with removing the latter.
--
Sakari Ailus
