On Thu, Aug 11, 2011 at 09:29:49PM +0200, Henrik Rydberg wrote: > The patch seems to do too many things - switching from static > variables to a dev struct, adding dev parameters to all functions, > changing to dynamic initialization, changing from platform to pnp > device. Two patches with no functional changes to start with, then > another two to turn on pnp seems reasonable. I've looked into this and I don't think it's especially practical to do it this way. As an example, applesmc_backlight_set calls applesmc_write_key, but it's a workqueue. The work struct is just a static structure, so it's not possible to obtain the device from there. It could just call the static platform device, but that just results in an intermediate stage that's worse than the previous one, which kind of defeats the object. If you'd like it broken up then the reasonable way would seem to be to do everything but the pnp transition in one patch and change to pnp in another. -- Matthew Garrett | mjg59@xxxxxxxxxxxxx _______________________________________________ lm-sensors mailing list lm-sensors@xxxxxxxxxxxxxx http://lists.lm-sensors.org/mailman/listinfo/lm-sensors
