Re: Enable shared lib tests for frv, lm32, m32r, microblaze, nds32 and or1k

[Kyrill Tkachov <kyrylo dot tkachov at foss dot arm dot com>]

Hi Alan,

On 07/12/17 09:50, Alan Modra wrote:
> On Wed, Dec 06, 2017 at 10:04:37AM +0000, Nick Clifton wrote:
> > Hi Alan,
> >
> > > +   * testsuite/lib/ld-lib.exp (check_shared_lib_support): Return true
> > > +   for frv, lm32, m32r, microblaze, nds32 and or1k linux targets.
> >
> > > -    && ![istarget lm32-*-*]
> > > +    && ![istarget lm32-*-elf]
> >
> > This introduced a whole slew of new failures when I tested a 
> lm32-rtems4.10
> > toolchain so I have checked in a little fix for this target:
> >
> > Cheers
> >   Nick
> >
> > ld/ChangeLog
> > 2017-12-06  Nick Clifton  <nickc@redhat.com>
> >
> >        * testsuite/lib/ld-lib.exp (check_shared_lib_support): Return
> >        false for lm32-rtems targets.
>
> Yes, I looked at that function and noticed it wasn't really doing the
> right thing.  I should have been a little less lazy and rewritten it.
>
>         * testsuite/lib/ld-lib.exp (check_shared_lib_support): Ask ld
>         under test whether -shared is supported.
>
> diff --git a/ld/testsuite/lib/ld-lib.exp b/ld/testsuite/lib/ld-lib.exp
> index 280a7b1..f960824 100644
> --- a/ld/testsuite/lib/ld-lib.exp
> +++ b/ld/testsuite/lib/ld-lib.exp
> @@ -1782,61 +1782,20 @@ proc check_gc_sections_available { } {
>  }
>
>  # Returns true if -shared is supported on the target
> -# Only used and accurate for ELF targets at the moment
>
>  proc check_shared_lib_support { } {
> -    if {![istarget aarch64*-*-elf]
> -        && ![istarget arc*-*-elf*]
> -        && ![istarget arm*-*-elf]
> -        && ![istarget avr-*-*]
> -        && ![istarget cr16-*-*]
> -        && ![istarget cris*-*-elf]
> -        && ![istarget crx-*-*]
> -        && ![istarget d10v-*-*]
> -        && ![istarget d30v-*-*]
> -        && ![istarget dlx-*-*]
> -        && ![istarget epiphany-*-*]
> -        && ![istarget fr30-*-*]
> -        && ![istarget frv-*-elf]
> -        && ![istarget ft32-*-*]
> -        && ![istarget h8300-*-*]
> -        && ![istarget i860-*-*]
> -        && ![istarget i960-*-*]
> -        && ![istarget ip2k-*-*]
> -        && ![istarget iq2000-*-*]
> -        && ![istarget lm32-*-elf]
> -        && ![istarget lm32-*-rtems*]
> -        && ![istarget m32c-*-*]
> -        && ![istarget m32r-*-elf]
> -        && ![istarget m6811-*-*]
> -        && ![istarget m6812-*-*]
> -        && ![istarget m68hc1*-*-*]
> -        && ![istarget mcore*-*-*]
> -        && ![istarget mep-*-*]
> -        && ![istarget microblaze-*-elf]
> -        && ![istarget mips*-*-elf]
> -        && ![istarget mn10200-*-*]
> -        && ![istarget moxie-*-*]
> -        && ![istarget msp430-*-*]
> -        && ![istarget mt-*-*]
> -        && ![istarget nds32*-*-elf]
> -        && ![istarget nios2-*-elf]
> -        && ![istarget or1k*-*-elf]
> -        && ![istarget pj-*-*]
> -        && ![istarget pru-*-*]
> -        && ![istarget rl78-*-*]
> -        && ![istarget rx-*-*]
> -        && ![istarget spu-*-*]
> -        && ![istarget v850*-*-*]
> -        && ![istarget visium-*-*]
> -        && ![istarget xc16x-*-elf]
> -        && ![istarget xgate-*-*]
> -        && ![istarget xstormy16-*-*]
> -        && ![istarget *-*-irix*]
> -        && ![istarget *-*-rtems] } {
> -       return 1
> +    global shared_available_saved
> +    global ld
> +
> +    if {![info exists shared_available_saved]} {
> +       set ld_output [remote_exec host $ld "-shared"]
> +       if { [ string first "not supported" $ld_output ] >= 0 } {
> +           set shared_available_saved 0
> +       } else {
> +           set shared_available_saved 1
> +       }
>      }
> -    return 0
> +    return $shared_available_saved
>  }
>
>  # Return true if target uses genelf.em (assuming it is ELF).

I'm seeing 202 unexpected failures on aarch64-none-elf with this patch.
With errors telling me to recompile with -fPIC or
"newlib/libgloss/aarch64/syscalls.c:639:(.text._sbrk+0x6c): dangerous 
relocation: unsupported relocation"
Should the tests perhaps be skipped on a bare-metal target?

Thanks,
Kyrill
> --
> Alan Modra
> Australia Development Lab, IBM