On Wed, Dec 13, 2023, at 10:58, Laurent Pinchart wrote: > On Wed, Dec 13, 2023 at 09:10:19AM +0000, Sakari Ailus wrote: >> On Wed, Dec 13, 2023 at 09:39:01AM +0100, Arnd Bergmann wrote: >> > On Wed, Dec 13, 2023, at 09:09, Sakari Ailus wrote: >> >> The datasheet says the input clock range is between 6 MHz and 54 MHz. We >> could check that, and return an error if it's not in the range. The rate is >> needed for other reasons, too, and the sensor is unlikely to work if it's >> (more than little) off. >> >> I wonder if this could address the Clang 16 issue, too? I'd replace >> udelay() with fsleep() in any case, and at the very least Clang should be >> happy with this. > > I'm fine with both of those changes. I verified that the build failure is solved by either one. I would just do the fsleep() change in that case since that is a sensible change regardless of the undefined behavior. I'll send a v2. Arnd