[DRBD-user] Re: New drbd .spec file for 0.6.10+
Lars.Ellenberg at linbit.com
Mon Mar 1 15:56:15 CET 2004
/ 2004-02-29 00:10:59 -0700
\ Sean Reifschneider:
> On Thu, Feb 05, 2004 at 10:12:33AM +0100, Philipp Reisner wrote:
> >Could you please confirm that, current CVS head has the .spec file
> >as it shoulde be ?
> It looks like the version in CVS doesn't allow installing both the SMP
> and non-SMP version. It also uses "km" instead of "kernel-module" for
> the prefix, and doesn't include the "drbd" name in the package name.
> Interesting choices for the name. I don't think I've ever seen that
> convention used before. I'd probably recommend at least putting "drbd"
> in the package name -- it's not like drbd is the only third-party module
> available. ;-)
It does, at least for SuSE. I don't know of the RH conventions.
According to my RPM documentation, the "main"-package name is prefixed
anyways, and thus when I do "make rpm" on my box for the typical
set of SuSE "preconfigured kernel headers", with this very spec
file, I get:
It is: "Main package name"-"km"-"kernel version" as subpackage name,
and the "drbd version" as version.
this is more natural to me, because the drbd.o ends up in
/lib/modules/"kernel version"/... anyways, so this may be
part of the subpackage name.
The SuSE convention for kernel modules is: they include it in
their binary kernel, and then have a binary km_<package>, which
basically is a minimal source package to "make ; make install",
in case you rebuild the kernel.
So I deviate from this, too, because I think if you need the
source and "make install" anyways, why don't you use the tgz.
Besides my binary drbd.o would not be what a SuSE user expects a
km_* package to be.
The suggested convention to have "kernel version" as part of the
version string is very confusing to me.
More information about the drbd-user