On Sun, 2007-04-29 at 14:31 +0100, David Woodhouse wrote:
At http://git.infradead.org/?p=users/dwmw2/livecd-ppc.git (git://git.infradead.org/~dwmw2/livecd-ppc.git) there's a set of patches which first introduces a little abstraction for the arch-specific configureBootloader() and createIso() functions, then implements the PPC version.
Cool -- there's definitely some abstraction needed, the question is more of where to put them. I think what might make sense is to get a few little things in the tree (like J5's fixes so that things work with FC6) and then branch off what we'll use for F7 and then continue to use HEAD so we can start doing more things like some of the refactoring from markmc as well as getting the architecture bits cleanly separated out.
I'll try to get to making progress on that this afternoon.
- Allow mayflower path to be specified, to run from git tree
I wonder if instead of specifying a path, it makes more sense to just notice that you're running from the git tree and use the mayflower script from there. The only problem is that sort of opens up to a certain class of attack. I'm not sure how important it really is to care about it, though.
2, Remove syslinux from package lists; add it dynamically instead.
The only problem with this is that it somewhat breaks the abstraction -- maybe some distro using livecd-tools needs the isolinux package for those bits. Does an error get raised with non-existent packages right now?
- Move x86-specific functions (createIso, configureBootloader) into subclass
- Only create isolinux/ directory in x86-specific configureBootloader()
- Add PowerPC support, add pata_mpc52xx module to mayflower.conf
This sounds pretty straight-forward and reasonable.
Jeremy