[PATCH, binutils] Fix implib test failures

[Thomas Preudhomme]

On Tuesday 19 July 2016 15:55:18 Nick Clifton wrote:
> Hi Thomas,
> 
>   Thanks for seeing this through.
> 
> > Is this ok for trunk?
> 
>   I do have a couple of points to make on the patch:
> 
> +# These targets use bfd_elf_final_link rather than bfd_generic_final_link
> +# and thus do not have support for ELF import library
> 
>   I think that you have this the wrong way around.  The targets are using
>   bfd_generic_final_link rather than bfd_elf_final_link and hence are
> missing the import library support.

huh, indeed.

> 
> +setup_xfail "am33lin-*-*" "d30v-*-*" "dlx-*-*" "i960-*-*" "m68hc1x-*-*"
> +setup_xfail "m88k-*-*" "pj-*-*" "score7-*-*" "sh64-*-*" "vxworks-*-*"
> 
>   There are other score targets apart from score7.  score-elf and score3-elf
> for example.  How about score*-*-* instead ?

I did not know how to get the affected triplet. All I got was a list of 
bfd/elf*-*.c that do not define elf_backend_relocate_section. I use the the 
value for the second * as the machine bit of the triplet. Is there other 
triplet in this list you see as not being inclusive enough? How can I find the 
list of triplet corresponding to a given bfd file?

> 
>   Also vxworks is not a cpu/architecture name, so there will never be any
>   targets that match vxworks-*-*.  Perhaps you meant: *-*-vxworks ?
> 
> 
> +     "--out-implib=tmpdir/implib.lib" ""
> 
>   I found that adding "-e 0" to the linker options helped reduce the
>   warning messages about the entry symbol not being defined.  It should
>   not make any difference to the tests - the harness ignores the warning
>   messages, but it makes the logs cleaner.

Will do when I respin the patch.

Best regards,

Thomas