On Sat, Dec 13, 2003 at 06:38:21AM -0500, Robert P. J. Day wrote: > not to sound like a complete idiot, but what's the rationale behind an RPM > not being relocatable? given the growing popularity of using linux for > embedded solutions, it would seem that creating a personal NFS-mountable > root filesystem on a server would make a lot of sense, and it would > definitely require relocatable RPMs to construct such a thing if you want > it to be RPM-based. obviously, non-relocatable RPMs would fail miserably > in this kind of situation. > You're confusing the definition of a relocatable package. For the purpose of your uses, you don't need a relocatable package, you simply want the --root functionality to do the chroot to your rootfs dir and then install there so you get the same behavior as what a native install would look like. You clearly just want to --initdb the --root and then install as you normally would with the --root specified. I find it curious why your install is succeeding at all with a --root specification if you haven't done the --initdb previously though. Regardless, the reason you are seeing your toplevel minicom config being touched is likely because you are getting -EPERM on the chroot(). For some reason, if rpm fails to do the chroot() to the --root path, it still continues on with the install silently without complaining loudly and bailing. You might want to run your install through strace and look for the chroot() call. If you are getting -EPERM, you'll likely have to do this as root (so glibc is probably not a good thing to start out with for testing ;-)
Attachment:
pgp00011.pgp
Description: PGP signature
