On Sun, 28 Nov 2004, Dave Jones wrote: > On Sun, Nov 28, 2004 at 01:07:58AM -0500, Matthew Miller wrote: > > > > ls /boot/config-`uname -r` > > > > I guess the argument is: there's a chance that something got skipped and the > > a kernel got built with the name matching a /boot/config-* file but actually > > with different options -- no risk of that with the /proc approach. > > There's no risk of it happening with the current approach either. > The .config that gets packaged is the .config rpmbuild faced > when it built the RPM. Why would it be different ? Well, I can imagine the situation that Fedora has already upgraded your kernel (read: replaced) and some clever mechanisme needs the kernel config to dynamically build a module for your current kernel. (DKMS?) Or when you have made your own kernel, and the config is not installed in /boot. Of course a tool can check that, can check the rpmdb and a lot of other things, but it would not have the same simplicity as a /proc/config.gz that is guaranteed to be there. (A lot of people rebuild kernels from the original Red Hat config file BTW) Again, I'm not in favor of one or another, but I can see some potential for having it around. One can disagree with that, but can't ignore that it gives more guarantees or potential. -- dag wieers, dag@xxxxxxxxxx, http://dag.wieers.com/ -- [Any errors in spelling, tact or fact are transmission errors]
