Packages failing build on PPC.

[David Woodhouse]

On Sat, 2005-02-05 at 11:28 +0000, David Woodhouse wrote:
> On Sat, 2005-02-05 at 11:51 +0200, Ville Skyttä wrote:
> > You don't happen to have the build logs available somewhere, so non-
> > PPC-owner package maintainers could have a look?
> 
> ftp://peach.ipv6.infradead.org/pub/SRPMS.extras/logs/

Also now http://david.woodhou.se/fc3-extras-ppc-logs/ for the benefit of
the IPv6-challenged (although most of you have no excuse for that).

I'd appreciate it if package maintainers could take a quick look -- if a
build log for your package exists, that's because it failed to build.

I can give accounts on PPC machines to package maintainers who need it.

Some things in there are OK to be i386-only, but others I'm not sure
about -- why sqlite, for example? I see PPC builds of that elsewhere. So
I've included all packages here.

In the long run, I'd like to see a policy of having to have a bug filed
to permit _any_ architecture exclusion. Even if that's a long-term thing
like "LILO doesn't make any sense at all on non-i386", I'd like to see
it filed, and I'd like to see the build system refusing to honour
ExcludeArch: or ExclusiveArch: without the bug number being listed in
the specfile. 

> gpgme fails its own gpgsm tests.
> 
> cvsup wants switching to cm3 which supports PPC. I may also separate
> modula3 and modula3-devel and a bunch of the libraries into separate
> packages. I did that once before, long ago.
> 
> Inventor needs a GCC bug fixed (the RHEL4 GCC also ICEs on this one). 
> 
> torcs seems to use -mieee-fp in its autocrap script and hence decides
> that libm isn't present on the system and aborts.
> 
> stow and gnugo fail due to /usr/share/info/dir being present in the
> buildroot but unpackaged.
> 
> fbida fails due to /usr/X11R6/lib/X11/app-defaults/Ida being present but
> unpackaged.
> 
> A whole bunch of other stuff fails due to random use of inline assembly.
> 
-- 
dwmw2