Re: [PATCH 11/15] [media] Deprecate v4l2_mbus_pixelcode

[Hans Verkuil <hverkuil@xs4all.nl>]

On 11/05/14 16:30, Boris Brezillon wrote:
> On Wed, 05 Nov 2014 16:19:56 +0100
> Hans Verkuil <hansverk@cisco.com> wrote:
> 
>>
>>
>> On 11/05/14 16:15, Boris Brezillon wrote:
>>> On Wed, 5 Nov 2014 17:08:15 +0200
>>> Sakari Ailus <sakari.ailus@iki.fi> wrote:
>>>
>>>> Hi Boris,
>>>>
>>>> On Tue, Nov 04, 2014 at 10:55:06AM +0100, Boris Brezillon wrote:
>>>>> The v4l2_mbus_pixelcode enum (or its values) should be replaced by the
>>>>> media_bus_format enum.
>>>>> Keep this enum in v4l2-mediabus.h and create a new header containing
>>>>> the v4l2_mbus_framefmt struct definition (which is not deprecated) so
>>>>> that we can add a #warning statement in v4l2-mediabus.h and hopefully
>>>>> encourage users to move to the new definitions.
>>>>>
>>>>> Replace inclusion of v4l2-mediabus.h with v4l2-mbus.h in all common headers
>>>>> and update the documentation Makefile to parse v4l2-mbus.h instead of
>>>>> v4l2-mediabus.h.
>>>>>
>>>>> Signed-off-by: Boris Brezillon <boris.brezillon@free-electrons.com>
>>>>> ---
>>>>>  Documentation/DocBook/media/Makefile |  2 +-
>>>>>  include/media/v4l2-mediabus.h        |  2 +-
>>>>>  include/uapi/linux/Kbuild            |  1 +
>>>>>  include/uapi/linux/v4l2-mbus.h       | 35 +++++++++++++++++++++++++++++++++++
>>>>>  include/uapi/linux/v4l2-mediabus.h   | 26 ++++----------------------
>>>>
>>>> I would keep the original file name, even if the compatibility definitions
>>>> are there. I don't see any harm in having them around as well.
>>>>
>>>
>>> That's the part I was not sure about.
>>> The goal of this patch (and the following ones) is to deprecate
>>> v4l2_mbus_pixelcode enum and its values by adding a #warning when
>>> v4l2-mediabus.h file is included, thus encouraging people to use new
>>> definitions.
>>
>> Since v4l2-mediabus.h contains struct v4l2_mbus_framefmt this header remains
>> a legal header, so you can't use #warning here in any case.
>>
> 
> Actually this patch moves the struct v4l2_mbus_framefmt definition into
> another header before adding the warning statement.

There is nothing wrong with keeping the struct in the header. Just keep it
there.

	Hans

> 
> Anyway, this is really a detail, and if everybody agrees that we should
> just leave the old definition in place, I'm fine with that.