Re: [REVIEW PATCHv1 1/2] DocBook: improve the error_idx field documentation.

[Laurent Pinchart <laurent.pinchart@ideasonboard.com>]

Hi Hans,

On Friday 11 January 2013 12:48:19 Hans Verkuil wrote:
> On Mon January 7 2013 20:56:07 Laurent Pinchart wrote:
> > On Monday 07 January 2013 13:09:47 Hans Verkuil wrote:
> > > From: Hans Verkuil <hans.verkuil@cisco.com>
> > > 
> > > The documentation of the error_idx field was incomplete and confusing.
> > > This patch improves it.
> > > 
> > > Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
> > > ---

[snip]

> > I think some flexibility should still probably be left to drivers (and I'm
> > not talking about UVC here), as some drivers might not be able to know
> > that a control is write-only before trying to read it and getting an
> > error.
>
> Well, if drivers don't know if a control is e.g. write-only until they try
> it, then it can't be done during pre-validation anyway, so that's no
> problem.

Sure, but my point is that we don't want to enforce in the spec that those 
checks must always be done during pre-validation, otherwise drivers that can't 
do it will violate the spec.

> The pre-validation should at minimum check whether ctrl_class is set up
> correctly, whether all controls in the list actually exist, and check
> against READ_ONLY or WRITE_ONLY (if known upfront).
> 
> The v4l2-compliance tool will test those minimum checks.
> 
> The control framework will also check whether the GRABBED flag is set for
> a control and if the new value of a control is valid.

-- 
Regards,

Laurent Pinchart