Re: [GIT PULL] at91: platform data for atmel-mci (for 3.5)

[Nicolas Ferre <nicolas.ferre@atmel.com>]

On 06/18/2012 03:49 PM, Olof Johansson :
> Hi,
> 
> On Wed, Jun 13, 2012 at 9:48 AM, Nicolas Ferre <nicolas.ferre@atmel.com> wrote:
>> On 06/04/2012 05:48 PM, Nicolas Ferre :
>>> On 05/31/2012 11:04 PM, Olof Johansson :
>>>> Hi Nicolas,
>>>>
>>>> On Thu, May 31, 2012 at 12:50 AM, Nicolas Ferre <nicolas.ferre@atmel.com
>>>> <mailto:nicolas.ferre@atmel.com>> wrote:
>>>>
>>>>     On 05/24/2012 05:12 PM, Nicolas Ferre :
>>>>     > Hi Arnd, hi Olof,
>>>>
>>>>     Ping?
>>>>
>>>>     (or maybe you will have a look at this after the merge window...)
>>>>
>>>>
>>>> So, I looked at this branch yesterday but I wasn't entirely happy with
>>>> the number of ifdefs it adds.
>>>
>>> Unfortunately, this is the usual shape of any devices/boards files on
>>> AT91. This amount will be reduced when we remove the old driver and when
>>> we move newer/popular devices to Device Tree...
>>>
>>>> If the idea is to deprecate the old driver, wouldn't it make more sense
>>>> to just cut everyone over instead of having both sets of setup in the
>>>> kernel?
>>>
>>> Well, the old driver has existed since a very long time and I think that
>>> people are used to it on oldest platforms. This is why we put in place a
>>> overlapping period. This way we hope that the transition will be smoother.
>>> On the other hand, the old code will be removed in 3.7 so the
>>> overlapping period will not be so long.
>>> I hope that it will allow people to track bugs if some are remaining and
>>> switch to newer driver easily.
>>
>> Olof, Arnd,
>>
>> So, do you have made up your mind about this pull request?
> 
> Sorry for the delays in handling this, I should have been quicker at replying.
> 
> That said, the driver is staged for removal in 3.7, and this patch is
> 3.6 material at this time. But I think it makes sense to cut every
> in-tree board over completely one release before the driver is
> removed, and thus not keep the old platform data around for them.
> 
> That way the out-of-tree users still have a one-release grace period,
> but everyone with an in-tree board (and the reference platforms) will
> move over sooner. I think I would prefer that over having all these
> ifdefs in the tree, even if it's just for one release.
> 
> I could be convinced otherwise if there's a good reason though. Either
> way, 3.6 is the way to go.

Fair enough, we will prepare a patch to remove the old platform data
combined with this patch. We will integrate in an "at91-3.6-soc"
series/branch that we will submit to you soon.

Thanks, best regards,
-- 
Nicolas Ferre