Created attachment 424628 [details, diff] Patch for /usr/portage/sys-libs/glibc/files/eblits/common.eblit See the attached patch... We don't really care whether the chost is sparc or sparc64 since neither of those give glibc enough information anyways. Multilib compilation is also broken with the existing logic as compiling 64 bit with a sparcv9 ctarget does not work. This patch merges the sparc64 / sparc logic and cheks $ABI / mcpu to see whether we should be targeting sparcv9/, sparcv8, or sparc64. If there's something better to use than $ABI let me know and ill fix it. This is required to get a pure sparc64 port and mulitlib port working.
Comment on attachment 424628 [details, diff] Patch for /usr/portage/sys-libs/glibc/files/eblits/common.eblit >+ if [ $ABI == "sparc32" ]; then Gentoo style is to use [[...]] and use braces for non-builtins. so this would read: if [[ ${ABI} == "sparc32" ]] ; then >+ cpu=sparc64v2 looks like you're mixing spaces & tabs here
Created attachment 424698 [details, diff] Revised glibc patch This patch should have those issues fixed.
Created attachment 425178 [details, diff] Revision 2 So i noticed that my catalyst builds since this patch were missing glibc, turns out i forgot to add the "if" in front for the statements... Not sure where my head was, so here's a corrected version. Interestingly enough the ebuild for glibc will still complete successfully with this error. It just wont compile glibc. I would think it should fail if there are syntax errors in an eblit...
Some comment from sparc would be nice here. A sparc of life!
eblits gone away and will need to be forward-ported to glibc ebuilds directly. Also, IFONC support appeared in sparc around 2010 and precise CPU selection might not be as relevant.
glibc dropped nearly all of these subconfigurations, since multiarch support made them unnecessary. See https://sourceware.org/git/?p=glibc.git;a=commit;h=5d9b7b9fa734c5381e0295c85c0e40520d9f6063 I've reworked the logic in the ebuilds in commit 7daa03609079248a30f64366345bd728e65da38e Author: Matt Turner <mattst88@gentoo.org> Date: Wed Mar 25 14:34:07 2020 -0700 sys-libs/glibc: Drop SPARC CHOST mangling Before "multiarch" support in glibc, picking a different CHOST would select which optimized paths were built. Multiarch support was added in glibc commit 3afd5a3b5556 (sparc: Add multiarch support for memset/bzero/memcpy.) in 2010 before glibc-2.12. Before glibc-2.30 glibc dropped support for SPARC v7 in commit 5d9b7b9fa734 (Remove 32 bit sparc v7 support) and in the process cleaned up some configuration logic that accepted a wide variety of (now unused CHOST values), thus causing the build to fail for us when we select a now unknown CHOST. Simply drop this logic, since it shouldn't be needed for any glibc since 2.11. Signed-off-by: Matt Turner <mattst88@gentoo.org> so I think we're done here.