On Thu, Apr 28, 2011 at 12:35 AM, Mark Brown broonie@opensource.wolfsonmicro.com wrote:
On Thu, Apr 28, 2011 at 12:17:46AM +0530, Jassi Brar wrote:
Please delete unneeded context, makes it much easier to find your text...
There wasn't anything to quote from changelog and my comments encompass your every change, so I commented at the end.
I have nothing against this approach, but just a thought... Is there any option like having 'dummy' versions of cpu, codec and dma dais, maybe ops too, and have machine drivers explicitly use them. That will preserve what has always been expected off of ASoC drivers and keep the core 'simple' (not to mean the checks you add above are serious overhead).
I considered that, my thought was mostly that it's better for the core to do a bit of work than for each user to have to do typing. I'd probably have gone the other way if we weren't already having to check to see if the op itself was there - this is why there's a dummy driver created and substituted in where there's no platform at all. Had I done the second patch first I'd probably never have written this one but it seemed like something someone might come up with another sane use case for some time.
In the DAIs we're actually substituting a dummy ops in but we shouldn't really be doing that as it ought to be possible for the drivers to be const.
OK.
Acked-by: Jassi Brar jassisinghbrar@gmail.com