rawhide report: 20090408 changes

Kevin Kofler kevin.kofler at chello.at
Thu Apr 9 22:29:12 UTC 2009


Jesse Keating wrote:
> OK fine, it needs to happen now.  Either way we can't go into the final
> freeze with a broken PPC tree.

PPC (and PPC64 even more so, but this particular issue is with 32-bit PPC)
really needs to become a secondary arch! This is not the first time PPC
issues are upholding builds. We've had several PPC/PPC64 issues with KDE
and its dependencies, and while it's true that some of them were due to a
generic bug (e.g. the CMake crash during kdepimlibs builds on ppc64 was
really a generic buffer overflow), most of them were due to PPC/PPC64
toolchain bugs (e.g. GCC crashing while building Qt - this is still not
fixed in the F10 GCC, by the way, but we were able to patch Qt so it
builds) or limitations (PPC64 TOC overflowing on large autogenerated
compilation units, we hit this with the OpenBabel SWIG bindings - upstream
ended up just removing some stuff from the bindings to "fix" the issue).
The Mono maintainers did the logical thing and decided to just drop PPC
support and I can't blame them for it. Only very few people still use PPC
hardware, all the current Macs are x86. What really needs to happen is for
PPC to become a secondary arch so the burden will be on the arch team to
get things to build there, maintainers should be able to move on with the
arches people actually use. ExcludeArch wouldn't have been needed then,
because the PPC builds would have been separate and could have been done on
their own schedule. This change has been promised for ages, why did it
still not happen?

        Kevin Kofler




More information about the fedora-devel-list mailing list