On Wed, Nov 16, 2005 at 04:19:58PM +0100, Jean Delvare wrote: <snip> > This is why you want to move the decision at the run time level rather > than the compilation time level whenever possible. The clock output > option can't be a configuration option. I previously suggested a sysfs > file, because this gives the greatest flexibility. If you don't like > the idea for whatever reason, you may go for a module parameter instead. > > Same reasonment holds for the m41t00 vs. m41t85 choice. You can't decide > at compilation time. If we go for a common driver, it has to support > both devices at the same time. Mark suggested to use platform-specific > data. I'm not familiar with this, but it sounds reasonable. I don't know the entire history behind platform_data but my understanding is that it was designed to provide a mechanism for platform-specific code to pass info to drivers. I *believe* that this would be a proper use of platform_data but I'm hoping someone out in lkml-land who knows better than I will confirm that. > I don't know for sure at this point whether having a single driver is > the right choice, I'll let you and Mark check it out and decide. But > the right way to determine this is definitely not through the use of > #if/#endif preprocessing stuff. I agree. We can and absolutely should do this at run-time if a merged driver is feasible. I'll dig thru the datasheets today and start prototyping some code if it looks like we can merge the code. Mark
