Le vendredi 11 décembre 2009 12:35, Michal Marek a écrit :
Jean Delvare napsal(a):
Le jeudi 10 décembre 2009 21:47, Michal Marek a écrit :
Jeff Mahoney napsal(a):
and mbuild is the only reason they're not. The build service can handle it. People using rpmbuild can use it. It's mbuild that's the issue. But is it practical then? I mean to build a kernel, you just build a kernel-$flavor.spec. If you make kernel-$flavor.spec dependent on kernel-source, then you always have to first build kernel-source.spec
... which is a sane thing to do anyway, as you "must" distribute kernel-source with any binary kernel package in order to comply with the GPL.
How does this relate to build ordering?
My personal experience is: given two thinks that must be done, one you like doing and one you don't like doing, chances to see the latter done are much higher if the former happens to depend on it in some way. Seriously, it doesn't strictly relate to the build ordering, but I don't think the build order really matters anyway.
(which spends good amount of time in rpmlint),
... peanuts compared to the build time of actual kernel flavors.
On a machine with lots of cpus, the non-paralelized parts like rpmlint and find-debuginfo.sh are the most annoying part. We try to optimize this, not add more.
If this really bothers you, maybe work on rpmlint and find-debuginfo.sh to let them parallelize their work? Oh, and I doubt find-debuginfo.sh takes that long for the kernel-source package ;) rpmlint, yes, certainly. -- Jean Delvare Suse L3 -- To unsubscribe, e-mail: opensuse-kernel+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-kernel+help@opensuse.org