Peter Stuge wrote: > On Sat, Jun 07, 2008 at 09:23:07PM -0700, Chris Kilgour wrote: > >> FYI - At this point I don't plan to submit any patches to gPXE as >> my approach is rather hackish and still experimental. >> >> > How do you think that a "proper" port of gPXE to coreboot differs > from your approach? > Off-hand I would say:
1. The gPXE elfprefix is currently busted, it would be preferred to fix that so gPXE might build a well-formed ELF on its own. At first I tried to fix it but gave up because of #2 below. 2. The "gPXE way" of producing final images is all about feeding a universal linker script, and I couldn't find a way of modifying it (trimming 16-bit code, getting the elfprefix right) to produce what I wanted without massive surgery that would affect a ton of regression cases. So I went the "treat gPXE as a library" route instead. 3. I'm not sure if all the 16-bit gPXE stuff really needs to be omitted or if it can/should be left in the ELF. I didn't want it, but others might. I also don't have the bandwidth to deal with getting a patch morphed into shape that the gPXE maintainers would accept, but I welcome anyone with such inclinations to take the baton. Chris. -- coreboot mailing list [email protected] http://www.coreboot.org/mailman/listinfo/coreboot
