Re: [RFC PATCH 2/3] v4l2-ctrls: add v4l2_ctrl_auto_cluster to simplify autogain/gain scenarios

[Hans de Goede <hdegoede@redhat.com>]

Hi,

On 01/23/2011 05:13 PM, Hans Verkuil wrote:
> On Sunday, January 23, 2011 16:15:03 Hans de Goede wrote:

<snip>

>> This is what the UVC spec for example mandates and what the current UVC driver
>> does. Combining this with an app which honors the update and the read only
>> flag (try gtk-v4l), results in a nice experience. User enables auto exposure
>> ->  exposure control gets grayed out, put exposure back manual ->  control
>> is ungrayed.
>>
>> So this new auto_cluster behavior would be a behavioral change (for both the
>> uvc driver and several gspca drivers), and more over an unwanted one IMHO
>> setting one control should not change another as a side effect.
>
> Actually, I've been converting a whole list of subdev drivers recently (soc_camera,
> ov7670) and they all behaved like this. So I didn't change anything.

Hmm, interesting.

> There is nothing preventing other drivers from doing something different.
>
> That said, changing the behavior to your proposal may not be such a bad idea.

Yes and AFAIK this is what we agreed on when we discussed auto control a
couple of months ago.

> But then I need the OK from all driver authors involved, since this would
> mean a change of behavior for them.
>
> The good news is that once they use the new framework function I only need
> to change what that function does and I don't need to change any of those
> drivers.
>
> So I will proceed for now by converting those drivers to use this new function,
> and at the same time I can contact the authors and ask what their opinion is
> of this change. I'm hoping for more feedback as well from what others think.
>

Yes, contacting the authors to discuss this further sounds like a good idea.

> BTW, if I understand the gspca code correctly then it seems that if an e.g.
> autogain control is set to 1, then the gain control effectively disappears.
> I think queryctrl will just never return it. That can't be right.

Erm, it should not disappear, but just get disabled. But this may have
(accidentally) changed with the drivers which were converted to the new
control framework. Anyways, we should discuss this with involved driver
authors and agree on a standard way to handle this. Once we have agreement
on how to handle this converting the drivers should be relatively easy.

Regards,

Hans