Alistair Crooks wrote: > On Sat, Jan 24, 2009 at 03:33:52PM +0100, Havard Eidnes wrote: > > > In its present form, the change is not complete, as the build fails > > > because the directory has not been created. > > > > > > nbmtree: ./kernel: No such file or directory > > > nbmtree: failed at line 35903 of the specification > > > > I've committed a fix to this by adding ./kernel to > > distrib/sets/lists/base/mi. Perhaps it should have gone into > > distrib/sets/lists/base/mi.module instead (I have a pending > > commit to etc/mtree/NetBSD.dist which adds ./kernel; without it, > > evbppc, which doesn't do kernel modules, will fail it's build). > > I believe that entries like a top-level "kernel" directory should go > into etc/distrib/NetBSD.dist, but I would much have preferred it if we > had reached consensus on where to put it before papering over the > cracks and fixing the build. I think the change should have been > reverted, an agreement reached, and then the change made. > > > For one, I support the renaming; ./stand is (at least historically) > > for stand-alone code, and kernel modules are not stand-alone code. > > Admittedly we have precious few, if any, instances of stand-alone code > > in our tree beyond our boot loaders, which are treated specially. > > "stand" may not be a good name for it - neither is "kernel". Something > else is necessary. I dont really think the name matters too much, just make it to be able to use different module dir for different kernels. Like if you boot netbsd.old make it use modules/netbsd.old/, so no symlink magic is required. -- When in doubt, use brute force. Adam Hoka <ahoka%NetBSD.org@localhost> Adam Hoka <ahoka%MirBSD.de@localhost> Adam Hoka <adam.hoka%gmail.com@localhost>
Attachment:
pgp2g1WkZNSPM.pgp
Description: PGP signature