Re: [RFCv2 PATCH 12/12] Remove audio.h, video.h and osd.h.

[Mauro Carvalho Chehab <mchehab@redhat.com>]

Em 25-11-2011 14:03, Andreas Oberritter escreveu:
> On 25.11.2011 16:38, Mauro Carvalho Chehab wrote:
>> Em 25-11-2011 12:41, Andreas Oberritter escreveu:
>>> On 25.11.2011 14:48, Mauro Carvalho Chehab wrote:
>>>> If your complain is about the removal of audio.h, video.h
>>>
>>> We're back on topic, thank you!
>>>
>>>> and osd.h, then my proposal is
>>>> to keep it there, writing a text that they are part of a deprecated API,
>>>
>>> That's exactly what I proposed. Well, you shouldn't write "deprecated",
>>> because it's not. Just explain - inside this text - when V4L2 should be
>>> preferred over DVB.
>>
>> It is deprecated, as the API is not growing to fulfill today's needs, and
>> no patches adding new stuff to it to it will be accepted anymore.
> 
> Haha, nice one. "It doesn't grow because I don't allow it to." Great!

No. It didn't grow because nobody cared with it for years:

Since 2.6.12-rc2 (start of git history), no changes ever happened at osd.h.

Excluding Hans changes for using it on a pure V4L device, and other trivial
patches not related to API changes, the last API change on audio.h and video.h 
was this patch:
	commit f05cce863fa399dd79c5aa3896d608b8b86d8030
	Author: Andreas Oberritter <obi@linuxtv.org>
	Date:   Mon Feb 27 00:09:00 2006 -0300

	    V4L/DVB (3375): Add AUDIO_GET_PTS and VIDEO_GET_PTS ioctls

	(yet not used on any upstream driver)

An then:
	commit 1da177e4c3f41524e886b7f1b8a0c1fc7321cac2
	Author: Linus Torvalds <torvalds@ppc970.osdl.org>
	Date:   Sat Apr 16 15:20:36 2005 -0700

	    Linux-2.6.12-rc2

No changes adding support for any in-kernel driver were ever added there.

So, it didn't grow over the last 5 or 6 years because nobody submitted
driver patches requiring new things or _even_ using it.

> 
>>>> but keeping
>>>> the rest of the patches
>>>
>>> Which ones?
>>
>> V4L2, ivtv and DocBook patches.
> 
> Fine.
> 
>>>> and not accepting anymore any submission using them
>>>
>>> Why? First you complain about missing users and then don't want to allow
>>> any new ones.
>>
>> I didn't complain about missing users. What I've said is that, between a
>> one-user API and broad used APIs like ALSA and V4L2, the choice is to freeze
>> the one-user API and mark it as deprecated.
> 
> Your assumtion about only one user still isn't true.
> 
>> Also, today's needs are properly already covered by V4L/ALSA/MC/subdev. 
>> It is easier to add what's missing there for DVB than to work the other
>> way around, and deprecate V4L2/ALSA/MC/subdev.
> 
> Yes. Please! Add it! But leave the DVB API alone!
> 
>>>> , removing
>>>> the ioctl's that aren't used by av7110 from them.
>>>
>>> That's just stupid. I can easily provide a list of used and valuable
>>> ioctls, which need to remain present in order to not break userspace
>>> applications.
>>
>> Those ioctl's aren't used by any Kernel driver, and not even documented.
>> So, why to keep/maintain them?
> 
> If you already deprecated it, why bother deleting random stuff from it
> that people are using?
> 
> There's a difference in keeping and maintaining something. You don't
> need to maintain ioctls that haven't changed in years. Deleting
> something is more work than letting it there to be used by those who
> want to.

Ok. Let's just keep the headers as is, just adding a comment that it is now
considered superseded.

>>> Btw.: It's not easy to submit a driver for a SoC. Even if you are
>>> legally allowed to do it, you have to first merge and maintain the board
>>> support code before even thinking about multimedia.
>>
>> Yes, I know that there's a long road for SoC drivers addition. Fortunately,
>> several vendors are now working to put their stuff upstream.
>>
>> I heard that there are some upcoming changes intended to simplify it a little bit,
>> trying to make the architecture a little more generic, and put board-specific
>> configuration on userspace. I dunno the details.
> 
> Thanks for your help.
> 
> Regards,
> Andreas