Re: Versioned symbols in elfutils and uClibc

[Mark Wielaard <mjw at redhat dot com>]

On Fri, Apr 24, 2015 at 07:49:55AM -0400, Anthony G. Basile wrote:
> On 04/23/15 18:24, Max Filippov wrote:
> >On Thu, Apr 23, 2015 at 11:56 PM, Mark Wielaard <mjw@redhat.com> wrote:
> >>On Thu, Apr 23, 2015 at 11:20:35PM +0300, Max Filippov wrote:
> >>>>What do the test results look like on your setup?
> >>>
> >>>I haven't tried that. Is there an easy way to run tests on a system
> >>>without native compiler?
> >>
> ># TOTAL: 126
> ># PASS:  44
> ># SKIP:  0
> ># XFAIL: 0
> ># FAIL:  82
> ># XPASS: 0
> ># ERROR: 0
> >
> >The log is attached.
> >A lot of tests failed because elfutils were built with --disable-progs
> >(otherwise the build fails), and some test binaries failed to build
> >because they call functions not available in uClibc.

Hmmm. elfutils doesn't support --disable-progs. Not having the tools
makes testing pretty impossible. I quickly looked over the log and it
does seem like all fails are indeed caused by the missing tools.

> You might also want to take a look a [1].  I had to produce a series of
> patches to get it to build.  What did you do about mtrace()?
> 
> [1] https://bugs.gentoo.org/show_bug.cgi?id=470884

Looking at that list it seems uclibc is missing a lot of support for
things elfutils relies on. I would really encourage to use a full
featured libc implementation like glibc. All testing on all arches is
done against it. That said, if people are really trying to make this
work lets see if we can support them a little. If the patches aren't
too intrusive and ugly we could apply them. But making sure they keep
working might be a pain because we don't have any regular testers on
such setups (on glibc based systems all supported arches should have
zero-fail).

Please see https://git.fedorahosted.org/cgit/elfutils.git/tree/CONTRIBUTING

Thanks,

Mark