Live image list in koji - SYSLINUX/ISOLINUX 6.02

[Kevin Kofler]

Kevin Fenzi wrote:
> It seems Kevin Koffler commited some changes to syslinux yesterday, I
> don't know if it fixes this issue, but that was his goal I think.

I did, but Peter Jones reverted my commit in dist-git and wants the live CD 
compose process changed to allow dragging in syslinux-nonlinux.i686 instead:
https://bugzilla.redhat.com/show_bug.cgi?id=1086446#c8

By the way, Mash did include syslinux-nonlinux-6.02-1.fc21.i686.rpm, 
syslinux-extlinux-nonlinux-6.02-1.fc21.i686.rpm and syslinux-
tftpboot-6.02-1.fc21.i686.rpm in the x86_64 repositories on the mirrors. The 
composes in Koji wouldn't find them anyway, and error out instead. (The 
x86_64 live image composes in Koji have all been failing due to that.)

Interestingly, my local x86_64 compose did complete, but failed to boot with 
the error message in #1086446, as did the i686 composes from Koji. So there 
seems to be more than one issue here. I don't know yet whether 6.02-2.fc21 
will work any better there, I'm waiting for it to hit the mirrors so I can 
respin my local live image compose.

It shall also be noted that the old syslinux 4.xx also included *.c32 files, 
they just weren't split into an i686-only subpackage, but shipped in the 
main syslinux.x86_64 (and also syslinux.i686) package. It's this duplication 
that Peter Jones wants to avoid. I disagree with him about the usefulness of 
those subpackages, and in particular on whether it's a worse hack to have 
some non-Linux (e.g. BIOS) 32-bit binaries (that neither BuildRequire nor 
Require any i686 packages) in an x86_64 package (as has been done in the 
past) or to require otherwise non-multilib x86_64 setups to ship i686 
packages to be able to boot (his subpackage setup).

        Kevin Kofler
        (with only one 'f' ;-) )