Re: [PATCH] [media] tvp5150: remove signal generator as input from the DT binding

[Rob Herring]

On Tue, Feb 23, 2016 at 08:28:14PM +0200, Laurent Pinchart wrote:
> Hi Javier,
> 
> On Tuesday 23 February 2016 15:23:48 Javier Martinez Canillas wrote:
> > On 02/23/2016 03:02 PM, Laurent Pinchart wrote:
> > > On Tuesday 23 February 2016 13:27:51 Javier Martinez Canillas wrote:
> > >> On 02/23/2016 01:16 PM, Laurent Pinchart wrote:
> > >>> On Tuesday 23 February 2016 13:09:58 Javier Martinez Canillas wrote:
> > >>>> The chip internal signal generator was modelled as an input connector
> > >>>> and represented as a media entity but isn't really a connector so the
> > >>>> driver was changed to use the V4L2_CID_TEST_PATTERN control instead.
> > >>>> 
> > >>>> Remove the signal generator input from the list of connectors in the
> > >>>> tvp5150 DT binding document as well since isn't a connector anymore.
> > >>>> 
> > >>>> Signed-off-by: Javier Martinez Canillas <javier@xxxxxxxxxxxxxxx>
> > >>>> 
> > >>>> ---
> > >>>> Hello,
> > >>>> 
> > >>>> I think is OK to change this DT binding because is only in the media
> > >>>> tree for now and not in mainline yet and also is expected to change
> > >>>> more since there are still discussions about how input connectors will
> > >>>> be supported by the Media Controller framework in the media subsystem.
> > >>> 
> > >>> I think that's fine, yes
> > >>> 
> > >>> Acked-by: Laurent Pinchart <laurent.pinchart@xxxxxxxxxxxxxxxx>
> > >> 
> > >> Thanks.
> > >> 
> > >>> I haven't noticed the patch that introduced this early enough I'm
> > >>> afraid, and I think we still have issues with those bindings.
> > >> 
> > >> Yes, I posted those patches and got merged before we had the discussion
> > >> about input connectors over IRC so I didn't know what was the correct way
> > >> to do it.
> > >> 
> > >>> The tvp5150 node should *not* contain connector subnodes, the connectors
> > >>> nodes should use the bindings defined in
> > >>> Documentation/devicetree/bindings/display/connector/ and be linked to
> > >>> the tvp5150 node using the OF graph bindings (ports and endpoints).
> > >> 
> > >> Agreed.
> > >> 
> > >>> Do you think you could fix that ?
> > >> 
> > >> Yes I will, I'm waiting for the input connectors discussions to settle so
> > >> I can post a final version of the DT bindings following what is agreed by
> > >> all.
> > >
> > > Shouldn't we revert the patch that introduced connectors support in the DT
> > > bindings in the meantime then, to avoid known to be broken bindings from
> > > hitting mainline in case we can't fix them in time for v4.6 ?
> > 
> > Yes, that would be a good idea. I've seen recently though a DT binding doc
> > that was marked as unstable / work in progress and I wonder if that's a new
> > accepted convention for DT binding docs or is just something that slipped
> > through review.
> 
> I'm not sure if it's an established practice but I certainly like it. However, 
> in this specific case, we know that the bindings are broken, so I think a 
> revert would be better.

It is not normal practice, though there are some exceptions.

> > The commit I'm talking about is f07b4e49d27e ("Documentation: bindings:
> > berlin: consider our dt bindings as unstable") but I don't see anything
> > documented in Documentation/devicetree/bindings/ABI.txt.

That one was simply not reviewed by anyone that would object. It's been 
almost a year, so they should be stable now...

Rob