gcc.dg/20000503-1.c added. was: reduced testcase for dwarf1 ICE

[Robert Lipe]

This sounded like approval.  I have committed the following as
gcc.dg/20000503-1.c.  The logs show the resulting ICE on both
i686-pc-linux-gnu and on i686-pc-sco3.2v5.  We might need to trim some
of the hosts later...

2000-05-03  Robert Lipe <robertlipe@usa.net>

        * gcc.dg/20000503-1.c: New test.

/* { dg-do compile { target i?86-*-linux* i?86-*-sco3.2v5* i?86-*-sysv* } } */
/* { dg-options "-O2 -gdwarf -g2" } */

/* Distilled from GCC's stmt.c.  Caused abort in dwarf-1 code. */

static void
expand_nl_goto_receiver ()
{
{
static struct elims {int from, to;} elim_regs[] = {{ 16, 7}, { 16, 6}, { 20, 7},{ 20, 6}};
      int i;

      for (i = 0; i < sizeof elim_regs / sizeof elim_regs[0]; i++)
        if (elim_regs[i].from == 16 && elim_regs[i].to == 6)
        break;
      }
}



Jeffrey A Law wrote:
> 
>   In message <20000502224641.K3029@rjlhome.sco.com>you write:
>   > This tiny test is inspired by a longstanding bootstrap failure on
>   > OpenServer.
>   > 
>   > Compile thusly 
>   > to watch it fail on an Linux/IA32 target:
>   > 
>   > $ ./xgcc -B./ -O2 -gdwarf -g2 ./d.c
>   > ./d.c: In function `expand_nl_goto_receiver':
>   > ./d.c:6: Internal compiler error in `dwarfout_file_scope_decl', at dwarfout
>   > .c:5307
> You should probably put this into the testsuite, probably gcc.dg and make
> it conditional on your target (if linux systems support dwarf1 debugging
> output, then enabling the test on those systems would be good too).
> 
> jeff