This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: [PATCH roland/bz14743] BZ#14743: Move clock_* symbols from librt to libc.
> sure, but no arches do this :). unless i missed something in your
> branch, if you were to merge/push that today, you'd break every arch
> except x86/x86_64/sparc/s390.
Every arch where the assembler groks %gnu_indirect_function but IFUNC
doesn't actually work, yes. That's what I said in the announcement. I
don't know which machines do or don't have working IFUNC and couldn't
quickly tell from inspection. That's why I left it to the port maintainers
for any machines lacking IFUNC to add the configure turd.
> true, but i don't think the two things are mutually exclusive. for all
> arches that lack ifunc support in the ldso, they should manually set
> libc_cv_asm_gnu_indirect_function=no. we should also have a test that
> makes sure the binutils supports it because all the ifunc support so far
> has been optional.
Sure. It's still optional, it's just that we newly actually need to know
whether it's usable or not. That would be the case too if someone added
any sysdeps/.../multiarch/ directories for a machine where there had ever
been a binutils version that passes the current %gnu_indirect_function test
but was not actually sufficient to build IFUNC-using objects.