Re: [PATCH] Fix strong_ref_weak_def.sh and dyn_weak_ref.sh when configured with --enabled-default-hash-style=gnu.

[Cary Coutant <ccoutant at gmail dot com>]

>>> >>> 2017-09-25  Romain Geissler  <romain.geissler@gmail.com>
>>> >>>
>>> >>>         * testsuite/Makefile.am (strong_ref_weak_def.sh, dyn_weak_ref.sh):
>>> >>>         Link with --hash-stye=sysv.
>>> >>>         * testsuite/Makefile.in: Regenerate.
>>
>> I'll stick my neck out and approve this.  I don't really have
>> authority to approve anything in gold/ but it looks near enough to
>> obvious to me.  We've needed to make the same sort of patch to fix
>> ld.bfd tests.
>
> Hi,
>
> Thanks for the approval, however I can't apply the patch myself since
> I have no write access to the repository. Can anyone apply it ?
>
> CCing Cary Coutant who is involved in gold develoment.

[Sorry, I try to keep an eye out for gold patches, but if you don't
put "gold" in the subject line or cc it directly to me, my email
filter won't highlight it for me, and I sometimes don't see the
message (or the many pings). I'll try to watch more closely.]

These tests are intended to check that the strong reference remains
strong, and shouldn't depend on which style hash table we're using. If
we generate the .stdout file with "readelf --dyn-syms -W" instead of
"readelf -sWD", we'll see the actual dynamic symbol table instead of
only what's in the hash table, and that's more what these tests should
be checking. I'd prefer to make that change rather than force the hash
style.

I'll also make a note to myself that there aren't any tests for
--hash-style. In particular, it seems that we should be testing for
this expected difference between sysv and gnu.

-cary