Hi, Ken -
> Thanks Frank for continuing to help out here ...
No problem, it's an annoying little mystery; thanks
for sticking with it!
> [...]
> $ rpm --eval %_target_cpu
> i386
Yup, that's where it starts going wrong.
> kenj@vm12:~$ uname -a
> Linux vm12 3.9.10-100.fc17.i686 #1 SMP Sun Jul 14 01:40:11 UTC 2013 i686 i686
> i386 GNU/Linux
Even on my 32-bit rawhide vm, that i386 bit at the end matches, so I
don't think that's a causal factor.
> I thought the VM config may have been (accidentally) changed at some
> point, and the CPU config had been something different at the
> initial install or at some subsequent boot or upgrade. So I blew
> the VM away, and started again with kvm style kvm32 Model CPUs
That was an excellent guess. Having looked through rpm's source code
briefly (the magic seems to be in lib/rpmrc.c), I backed away with the
sweat of fear, and remain mystified why it'd have chosen i386 for you.
> ... reinstalled Fedora 17 32-bit and all the needed RPMs, recloned
> the PCP tree, but Makepkgs dies the same way and the /etc/rpmrc
> "hack" does provide a workaround [...]
Thanks. For better or for worse, F17 has already lost upstream
support, as has F18 for that matter. F19 and F20 are supported
(and have libatomic, so the other build fix would kick in).
(Another option for building distro-flavoured RPMs is to use mock to
maintain & use hygienic rpm buildroot environments.)
- FChE
|