Unmodified, eselect-python cannot be installed on IRIX (unless using GCC) - seemingly needlessly. The following change to the ebuild will allow it to work correctly: --- app-admin/eselect-python/eselect-python-20091230.ebuild +++ app-admin/eselect-python/eselect-python-20091230.ebuild @@ -18,19 +18,19 @@ IUSE="" RDEPEND=">=app-admin/eselect-1.2.3" DEPEND="${RDEPEND} sys-devel/autoconf - >=sys-devel/gcc-3.4 dev-libs/gnulib" pkg_setup() { - if [[ $(gcc-major-version) -lt 3 || ($(gcc-major-version) -eq 3 && $(gcc-minor-version) -lt 4) ]]; then - die "GCC >=3.4 is required" - fi if [[ ${CHOST} == *-solaris2.9 ]] ; then # solaris2.9 does not have scandir yet append-flags -I"${EPREFIX}/usr/$(get_libdir)/gnulib/include" append-ldflags -L"${EPREFIX}/usr/$(get_libdir)/gnulib/$(get_libdir)" append-libs -lgnu fi + + if [[ ${CHOST} == *-irix* ]] ; then + append-cppflags -D__attribute__\\\(x\\\)=\"\" + fi } src_unpack() { ... although further integration work will be needed to bring back the GCC tests where this is the main compiler.
we better fix that wrapper so it doesn't need any tricks to compile
I think I fixed this mess in -r00.1, can you check?
Great - the -r00.1 version build just fine! IRIX-MIPSpro-wrapper will inject the necessary define into all C*FLAGS if the environment variable MIPSPRO_INJECT is set - but this is a pretty blunt instrument. Much better is if developers guard GNU-specific constructs from non-GNU compilers - which is why the option isn't enabled by default. To be fair, most code does get this right...
yeah, it's very stupid code we're talking about here. For the next bump I'll fork it for Prefix, since "upstream" is unable to understand this needs to be done differently.
no info needed