[alsa-devel] [PATCH 04/20] ASoC: S3C: I2Sv2: Remove S3C_IIS_V2_SUPPORTED define

[jassi brar]

On Wed, Mar 10, 2010 at 7:49 PM, Mark Brown
<broonie at opensource.wolfsonmicro.com> wrote:
> On Wed, Mar 10, 2010 at 04:48:53PM +0900, Jassi Brar wrote:
>> The code in s3c-i2s-v2.c is compiled via Makefile and not implicitly
>> by include'ing. Also, the driver developer anyways has to refer to the
>> manual to see if the code can be reused for the SoC under consideration.
>> That makes the S3C_IIS_V2_SUPPORTED retrospective rather than a check.
>
>> Signed-off-by: Jassi Brar <jassi.brar at samsung.com>
>
> This code is in there for pure defensiveness - since at least the
> S3C64xx and S3C24xx have slightly different register maps for the block
> (the ones that are currently handled are the IISMOD master settings,
> there may be more) the code needs to have explicit support for the
> processors.  The goal of this block is to print a #error saying exactly
> what's going on rather than just have the build fail with missing
> defines in order to help make it more immediately clear to users what is
> going on if they try to select a processor combination they don't know
> how to handle.
The way I see it, if we keep it we will have to add every supported
SoC to the list
which is kinda redundant since that information could already be extracted from
Makefile. The list of supported SoC could be useful if we somehow knew
the future SoCs' characteristics to tell if they wud work or not, but
the developer
has to first check the Manual and see if the s3c-i2s-v2.c could be
made use of or not.
Besides, I assume 'user' to be the MACHINE driver writer and 'developer' to be
the CPU driver writer.

> Users often have a strong expectation that the driver will just work
> and so relying on them consulting the processor manual isn't always
> realistic.
well, I am not sure if we should sympathize with such a developer :)

> It may be that these differences are due to some being IISv2 and some
> being IISv3 or something and could easily be handled via mandatory
> platform data but since none of the documentation refers to any of this
> stuff and it has to be reverse engineered we've got this system based on
> build time processor selection.
Yes, maybe some more documentation is what is needed.