On Mon, May 27, 2024 at 08:03:12PM +0200, Niklas Söderlund wrote: > On 2024-05-27 17:37:21 +0100, Conor Dooley wrote: > > Should be with the driver: > > https://lore.kernel.org/all/20240527132429.1683547-1-niklas.soderlund+renesas@xxxxxxxxxxxx/ > > As I mentioned in the other thread about the ISPCS bindings, I > intentionally posted the bindings separately to allow parallel > upstreaming of driver and DT users. > > Is it really a bad idea to do it this way? For other work I have done > that involves more complex DT changes then adding a compatible, such as > adding a new device or adding more properties to cover more features > only available in a later version of a device. I always post the DT > parts first as this can spur discussions about the design and only after > they are agreed upon do I post the driver parts that make use of them. > > Seems like this would consume less review resources as the bindings can > be agreed upon first, before anyone have to spend time reviewing a > driver that might need to be redesigned as the bindings could be > improved. I would always rather have the driver implemented rather than just discuss some idealised bindings. The first place I go when I have concerns or confusion about how a binding is intended to be used is the driver - so it doesn't make my time reviewing something easier, that's for sure. Additionally, having a software implementation can make it obvious where mistakes may be in a complex binding or notice things that were omitted. Getting a binding merged early in that case can easily become a hinderance..
Attachment:
signature.asc
Description: PGP signature
