On 26.08.2010 13:18, Marc Kleine-Budde wrote: > Bhupesh SHARMA wrote: >> I am in parallel preparing the patch against david miller's 2.6.35 and I >> will try >> to get it pushed in the netdev-next mailing list in the coming days. >> However, I wished >> to get some review comments from this list before submitting the same, >> mainly because the >> reference of this driver is the old bosch CCAN driver present in the SVN >> tree. >> >> That is why I added the LINUX_VERSION_CODE stuff. > > I personally don't care about the svn (for drivers). net-next-2.6 is the > reference. IMHO writing good kernel code is hard enough and adding > ifdefs for older Kernels doesn't make it easier. It's not maintain- nor > testable. > > If someone needs this driver for legacy kernels he/she can add the > appropriate ifdefs and test it with that old kernel and get the driver > into the svn.
Yes, that's also my impression about the best workflow with new drivers. The process to get drivers into mainline became pretty fast - especially due to the valuable CAN-specific feedback on this ML :-) To maintain drivers in the SocketCAN SVN is an extra effort. That's why the mcp251x.c branched unfortunately. Btw. it's always nice when users give back a backported code for other people having the same problems with older kernels. I assume the importance of the kernel code in the SVN is decreasing with every new kernel version hitting the road. Regards, Oliver _______________________________________________ Socketcan-core mailing list [email protected] https://lists.berlios.de/mailman/listinfo/socketcan-core
