Re: RFA/RFC: insns that do not start a source line

[Nick Clifton]

Hi Bernd,

Could you point me at the code generating these NOPs

It is in cfgrtl.c:fixup_reorder_chain()


      nb = split_edge (e);
      if (!INSN_P (BB_END (nb)))
        BB_END (nb) = emit_insn_after_noloc (gen_nop (), BB_END (nb),
                                                         nb);
      INSN_LOCATION (BB_END (nb)) = e->goto_locus;

 and maybe a testcase)?

Full testcase attached.  Compile with -O0 -g.

 Also, is this the only case you know of or are there others?

Sorry, I have not investigated this.  The problem was reported to me by 
a GDB engineer, who was investigating gdb's failure to apply breakpoints 
to the correct locations under some circumstances (such as the code in 
the testcase).  There may well be other tests that have similar 
properties, especially loops that are unrolled or goto statements that 
cross basic blocks.

Cheers
  Nick


volatile int v;
volatile int w;

void
loop_test (void)
{
  v = 0;

  while (v < 3)                         /* Loop 1 condition */
    {
      v++;                              /* Loop 1 increment */
    }

  v = -42;

  for (v = 0; v < 3; )                  /* Loop 2 */
    {
      v++;                              /* Loop 2 increment */
    }

  v = -42;
  w = 42;

  for (v = 0;                           /* Loop 3 initialization */
       v < 3;                           /* Loop 3 condition */
       v++)                             /* Loop 3 increment */
     {
      w = w - v;        
     }

  v = 0;
  goto b;                               /* Loop 4 initial goto */
a:  v++;
b:  if (v < 3) goto a;                  /* Loop 4 condition */
}

int main (int argc, char **argv)
{
  loop_test ();

  return 0;
}