On Tuesday 17 April 2012 20:06:19 Takashi Iwai wrote:
At Tue, 17 Apr 2012 18:04:31 +0100,
Mark Brown wrote:
On Tue, Apr 17, 2012 at 06:52:49PM +0200, Takashi Iwai wrote:
Mark Brown wrote:
This isn't just adding something into a specific driver which fails at abstraction, it's adding generic code. If it were adding something to the ice17xx driver then that'd be one thing but look at the subject line and location of the file... this stuff should be buried inside the driver if it's too painful to make the driver sane.
The codes in sound/i2c are mostly oly for ice1712/ice1724 drivers after all... They could be used by others, but I don't think there will be any more at this point.
If they're specific to that driver we should make them specific to that driver and make sure the pain is confined there. We really don't want to end up going back to the bad old days of having to do per-CPU/card drivers for CODECs because nobody had thought to abstract this stuff, that just makes everyone miserable.
Looking at these commits I'd not expect anyone to figure out that this isn't how we want or expect people to add generic CODEC drivers.
Yeah, these stuff can be better put in pci/ice1712/ directory only for ice1724 driver. In that way, you can avoid unnecessary exported symbols, too.
Xonar (oxygen) has a private version of these drivers too. It could be converted to use common code instead.