On 9/3/07, Peter Chubb <[EMAIL PROTECTED]> wrote: > I usually use Debian, but for a project I need to work in a > Redhat or Centos-like environment. So I decided to create a > CentOS chroot. > > I did this by creating a largeish virtual disk for qemu, and > then booting qemu from the first installaton disk. I can then > later loopback mount the virtual disk and copy the files into a > directory for chrooting to. But CentOS seems to want to > install lots of guff I don't want. It also wants to instal > stuff from disks other than the first. How do I get a truly > minimal installation -- just enough to run yum ? Is there an > equivalent to debootstrap for CentOS/Fedora/RedHat ?
Years ago when we both worked at the place that no longer exists, I found it took about 50 rpms in a chroot to get to the place where I could rebuild a kernel from a source rpm. I seem to remember finding that number quite obscene at the time. I'm pretty sure I used rpm to install stuff into the chroot, repeatedly adding dependencies on the command line until I had a set that was complete. I probably had a --bind mount inside the chroot containing the .rpm files off the relevant install disks, so the actual install was done from inside with no media exchanges required. I have vague memories that the rpm binary may have been statically linked to avoid issues with needing all sorts of stuff already there to make it work. Since I did the work for hire, I didn't feel I could take a copy of my script away for times like today. :-( Putting that aside, there is supposedly an 'rpmstrap' thingy, but I have no idea whether it's worth the effort. As my Linuces these days are not .rpm based, I can't say too much more, other than 'best of luck'. This all reminds me to be thankful I can install my preferred OS over a network using a PXE booted kernel and no physical media. -- Christopher -- SLUG - Sydney Linux User's Group Mailing List - http://slug.org.au/ Subscription info and FAQs: http://slug.org.au/faq/mailinglists.html
