Hello Laurent, On Wed, Mar 18, 2020 at 12:56:56AM +0200, Laurent Pinchart wrote: > On Tue, Mar 17, 2020 at 02:32:27PM +0200, Oleksandr Suvorov wrote: > > PWM can have a normal polarity and a reverted one. The reverted polarity > > value is defined. > > I would squash this patch with 2/7, apart from that it looks fine. > However, I also agree with Thierry that the PWM cell that contains this > value is a bitmask, so once we get more flags it may get a bit awkward. For me the usefulness of PWM_POLARITY_NORMAL increases with more bits used. That's because if there are 5 things that can be set there and the patch author mentions only the two that are non-zero, I as a reviewer don't know if the author actually know and thought about the other three. If however they spell out PWM_POLARITY_NORMAL it's quite sure they want normal polarity. > Will we have one macro for each flag that will evaluate to 0 to report > that the flag isn't set ? Yes. Given the above mentioned advantage this is cheap enough in my eyes. > Or should we define a single PWM_FLAG_NONE (or > similarly named) macro ? I like one macro for each bit field better for the above mentioned reason. > In retrospect, maybe PWM_POLARITY_INVERTED > should have been named PWM_FLAG_POLARITY_INVERTED. Seems to be subjective. I don't see much added semantic that justifies the longer name. Best regards Uwe -- Pengutronix e.K. | Uwe Kleine-König | Industrial Linux Solutions | https://www.pengutronix.de/ |
