Re: [PATCH V1 1/3] ASoC: fsl: Kconfig: remove dependence of SND_IMX_SOC for SND_SOC_IMX_PCM_DMA

[Lars-Peter Clausen <lars@metafoo.de>]

On 08/19/2014 09:41 AM, Shengjiu Wang wrote:
> On Mon, Aug 18, 2014 at 06:52:46PM +0200, Lars-Peter Clausen wrote:
>> On 08/18/2014 10:38 AM, Shengjiu Wang wrote:
>>> Build kernel with SND_SOC_IMC_PCM_DMA=m && SND_IMX_SOC=n leads the following
>>> error:
>>>
>>>     sound/built-in.o: In function `fsl_sai_probe':
>>>>> fsl_sai.c:(.text+0x5f662): undefined reference to `imx_pcm_dma_init'
>>>     sound/built-in.o: In function `fsl_esai_probe':
>>>>> fsl_esai.c:(.text+0x6044b): undefined reference to `imx_pcm_dma_init'
>>>
>>> Most cpu driver in soc/fsl has use the function 'imx_pcm_dma_init' which is
>>> defined in imx-pcm-dma.c, so need to select SND_SOC_IMX_PCM_DMA, but it depends
>>> on SND_IMX_SOC, if SND_IMX_SOC=n, then SND_SOC_IMX_PCM_DMA will not be
>>> selected.
>>> So remove the dependence of SND_IMX_SOC for SND_SOC_IMX_PCM_DMA.
>>>
>>> Reported-by: kbuild test robot <fengguang.wu@intel.com>
>>> Signed-off-by: Shengjiu Wang <shengjiu.wang@freescale.com>
>>
>> Those if conditions where just added[1] by Arnd to avoid other build
>> failures. So just removing them again makes little sense.
>>
>> As far as I can see imx_pcm_dma_init() should be stubbed out when
>> SND_SOC_IMX_PCM_DMA is not selected. So what's going on here?
>>
>> [1] http://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/?id=ff40260f79dc0436604452bccd449bffd25ebafb
>>
> Lars-Peter Clausen & Arnd
>
> The purpose of Arnd's patch is same with me, which is to resolve the build
> error when SND_SOC_IMX_PCM_DMA=m & SND_SOC_FSL_SSI/SAI/ESAI/SPDIF=y, the
> error is "undefined reference to `imx_pcm_dma_init'".
> But Arnd's patch didn't involve this situation that SND_IMX_SOC=n &
> SND_SOC_IMX_PCM_DMA=m.

Having SND_SOC_IMX_PCM_DMA selected if SND_IMX_SOC is not selected makes no 
sense. I think the proper fix is to find out why it is selected and make 
sure that it is not selected when SND_IMX_SOC is not selected.

>
> Currently I think out a solution for this issue is to change
>
> select SND_SOC_IMX_PCM_DMA if SND_IMX_SOC != n
>
> to
>
> depends on SND_SOC_IMX_PCM_DMA != m
>
> How do you think about this?

No, that has the dependencies in reverse.