execution test fails - XFAIL powerpc-*-eabi

[mrs@wrs.com (Mike Stump)]

To the author of execution test fails - XFAIL powerpc-*-eabi in
g++.mike/dyncast[12].C, it doesn't do what you think it does.  You
have to add support for XFAIL, if you want to use it.

But caution, I tried to add it for you, and got caught in the old
semantics games.  What we appear to want, is a execution test - XFAIL
powerpc-*-eabi in this case, and then add support for that.  The
reason is we don't want universal complement target triplets in the
common case (because we don't have them), and we want execution test
fails - to, by default, setup an expected failure for all machines.

I think we should do something like the below in the meantime.  This
gets rid of the xpass for those two cases on most machines.

Doing diffs in testsuite:
*** testsuite/g++.old-deja/g++.mike/dyncast1.C.~1~	Wed Oct 29 16:07:54 1997
--- testsuite/g++.old-deja/g++.mike/dyncast1.C	Wed Nov  5 18:09:42 1997
***************
*** 1,6 ****
  // Special g++ Options: -fexceptions -w
  // excess errors test - XFAIL a29k-*-* sparc64-*-elf sh-*-* arm-*-pe
- // execution test fails - XFAIL powerpc-*-eabi
  
  #include <typeinfo>
  
--- 1,5 ----
*** testsuite/g++.old-deja/g++.mike/dyncast2.C.~1~	Wed Oct 29 16:07:55 1997
--- testsuite/g++.old-deja/g++.mike/dyncast2.C	Wed Nov  5 18:09:45 1997
***************
*** 1,6 ****
  // Special g++ Options: -fexceptions -w
  // excess errors test - XFAIL a29k-*-* sparc64-*-elf sh-*-* arm-*-pe
- // execution test fails - XFAIL powerpc-*-eabi
  
  // Ensure reference handling works.
  
--- 1,5 ----
--------------