Re: is_underscore_target for ld-elf tests

[Alan Modra <amodra at gmail dot com>]

On Sun, Jun 18, 2017 at 11:47:37PM -0400, Hans-Peter Nilsson wrote:
> > -  cris-*-* | crisv32-*-*)
> > +  cris-*-*)
> > +    targ_defvec=cris_aout_vec
> > +    targ_selvecs="cris_elf32_us_vec cris_elf32_vec ieee_vec"
> > +    targ_underscore=yes
> > +    ;;
> > +
> > +  crisv32-*-*)
> >      targ_defvec=cris_aout_vec
> >      targ_selvecs="cris_elf32_us_vec cris_elf32_vec ieee_vec"
> > -    targ_underscore=yes # Note: not true for cris_elf32_vec.
> >      ;;
> >
> 
> No, this is just as wrong.  crisv32-* and cris-* are the same
> wrt. underscoreness: true for *-elf (and *-aout which nobody
> uses), false for *-linux-gnu.  I don't remember why I had it
> set uniformly to yes.  Maybe way back, it didn't matter.

Oops, badness crept in because I regression test cris-elf and
crisv32-linux, and between reading gas/config/tc-cris.c and finding
out exactly what was going on, and making the final patch I confused
the issues.  (The first patch hacked is_underscore_target for cris,
then I figured I ought to make c++filt work properly for cris too.)

I'll commit the following after a quick test.

diff --git a/bfd/config.bfd b/bfd/config.bfd
index f8a26f4..dc24aab 100644
--- a/bfd/config.bfd
+++ b/bfd/config.bfd
@@ -500,15 +500,13 @@ case "${targ}" in
     targ_underscore=yes
     ;;
 
-  cris-*-*)
-    targ_defvec=cris_aout_vec
-    targ_selvecs="cris_elf32_us_vec cris_elf32_vec ieee_vec"
-    targ_underscore=yes
-    ;;
-
-  crisv32-*-*)
+  cris-*-* | crisv32-*-*)
     targ_defvec=cris_aout_vec
     targ_selvecs="cris_elf32_us_vec cris_elf32_vec ieee_vec"
+    case "${targ}" in
+	*-*-linux*) ;;
+	*) targ_underscore=yes ;;
+    esac
     ;;
 
   crx-*-elf*)

-- 
Alan Modra
Australia Development Lab, IBM