On Wed, Jan 21, 2009 at 12:11:29AM +0900, YAMAMOTO Takashi wrote: > hi, > > > On Mon Jan 19 2009 at 23:25:40 +0000, Andrew Doran wrote: > > > > +#ifdef COMPAT_50 > > > > +#include <compat/sys/cpuio.h> > > > > +#endif > > > > > > > > This doesn't work properly any more, when put in code outside a compat > > > > module, because i386 (soon x86) doesn't get this in GENERIC. I sent > > > > a mail about it a while ago, can't remember the title. > > > > > > As an aside, the documented interface is cpuctl(8). It would be good to > > > have > > > a policy on compatibility (my vote, obviously, is application programs > > > only!). > > > > I discussed this with Christos and a few others in passing earlier today, > > and your suggestion is my vote too. Providing compat is nice, but I don't > > think it's worth the effort for every foobar ioctl and knob in the world > > to benefit the two people who are too lazy to upgrade their whole system. > > i personally fully agree that we should relax the compatibility policy. While I understand your position, I'm not sure how you can relax it only by a little. If we want a consistent policy in the long run, it will either be what we do today, fighting teeth and nails so that binary compatibility is kept (and sure enough, we don't always succeed, but we do have a fairly good record), or we just keep it at the bare minimum so that the user can do "build.sh install=/; shutdown -r now" with a newer kernel. Anything in-between will always lead to long discussions, and I think we already have enough of those already. -- Quentin Garnier - cube%cubidou.net@localhost - cube%NetBSD.org@localhost "See the look on my face from staying too long in one place [...] every time the morning breaks I know I'm closer to falling" KT Tunstall, Saving My Face, Drastic Fantastic, 2007.
Description: PGP signature