On Sun, 2006-10-22 at 10:01 -0500, Johnny Hughes wrote: > Yum does not install kernel-smp-devel (or kernel-hugemem-devel, kernel- > largesmp-devel), but it upgrades them. > > Yum does install kernel-devel, and it does not upgrade it. > > Yum should be consistent in how kernel-devel files are handled. > > ----------------------------------------------- > > This issue is addressed in this CentOS bug: > <snip bug refs, etc.> > There are 3 options here: > > 1. Patch CentOS-4 yum to make yum install all kernel-*-devel files like > it does for kernel-devel (or the reverse ... make kernel-devel and > upgrade like the other files). > > 2. Modify the kernel-2.6.spec to do what FC >= 4 does. > > 3. Do nothing and tell people to choose what they want by updating this > variable in /etc/yum.conf > > installonlyn= > > ------------------------------------ > > I think that 2 is a bad choice as it makes the CentOS kernel deviate > from the upstream kernel. > > I think that 3 is a better choice than 2 ... but I think 1 is the > optimal choice. That yum needs to be updated to treat kernel-*-devel > files like it treats kernel-devel. > > What does everyone else think? I almost always vote for the solution indicated by "rule of least surprises". What is that in this case? Depends on who you are what your background is? I would *guess* the fewest "support requests" would occur if the installation of any *dev* for the kernel installed any needed predecessors. Installations and upgrades should then track. Since upstream doesn't do yum, no concern about breakage? I don't think there should be any gripes if the CentOS-specific install/upgrade process followed its own path. It's not really deviating from upstream in what is available/delivered any more than using YUM already diverges. It's just doing a little better job of being consistent in the installation/upgrade process. If the spec is modified to accomplish this, it just means the bug is fixed in CentOS a little earlier and all CentOS users and support folks benefit. But that seems a more "serious" divergence as you now have to maintain that difference from upstream. Not so with yum: there is no equivalent upstream? Fewest surprises there. But I'm basically ignorant of all the ramifications and details. > ----------------------------------- > Note: This is my attempt to solicit input for package changes from the > public and not make unilateral decisions and push them with only the > developers present. I didn't think Texans caved so easily! ;-) > > If there is no discussion of this item on this list by non-centos > developers, then I will revert back to making bug changes based on only > what the developers think :P They do?! :=)) > <snip> > Johnny Hughes > <snip sig stuff> -- Bill