At Thu, 5 Jul 2012 10:28:42 +0100, Mark Brown wrote:
On Thu, Jul 05, 2012 at 09:06:30AM +0200, Takashi Iwai wrote:
Mark Brown wrote:
There is no need for the constraints to be modified while being applied
Many drivers pass the local instance to rule->private, and it's not always guaranteed to be const. For example, you can imagine some state modified kept in the struct while it's modified via rules.
In short: you'd have seen many compile warnings for non-ASoC drivers if you run make once with this patch ;)
Well, that's trivial to fix.
As a safe side fix, how about just adding const to snd_pcm_hw_constraint_list() & co, and expclitly cast to non-const later?
My inclination for something like this is to put the casts with the users since that way we get the type safety through most of the code and we can see that dropping the const is safe since we're just getting back our own data. If we can see the APIs immediately dropping the const I'd expect we'd get people spending time on code review trying to figure out if it's safe.
The rule->private pointer isn't designed to be always const. The data purely depends on each rule function, thus whether it can be const or not also depends on the rule, too. For simple rules like list take indeed the constant list. But more complex rules may take a struct containing variables changed on the fly (e.g. the max rate set by user via control API).
Takashi