[Bug gdb/29962] FAIL: gdb.reverse/step-reverse.exp: reverse step into fn call

["vries at gcc dot gnu.org" <sourceware-bugzilla@sourceware.org>]

https://sourceware.org/bugzilla/show_bug.cgi?id=29962

--- Comment #4 from Tom de Vries <vries at gcc dot gnu.org> ---
(In reply to Tom de Vries from comment #3)
> As mentioned before, this failure shows up with gcc 9.4.0.
> 
> It doesn't show up with gcc 12.2.1, because of a difference in generated
> code.
> 
> In the 12.2.1 case we have 4 insns:
> ...
>   401208:       8b 55 cc                mov    -0x34(%rbp),%edx
>   40120b:       8b 45 d0                mov    -0x30(%rbp),%eax
>   40120e:       29 c2                   sub    %eax,%edx
>   401210:       89 55 d4                mov    %edx,-0x2c(%rbp)
> ...
> each with their line number entry:
> ... 
> step-reverse.c         78            0x401208               x
> step-reverse.c         78            0x40120b               x
> step-reverse.c         78            0x40120e               x
> step-reverse.c         78            0x401210               x
> ...

In the gcc 9.4.0 case, we have 5 insns:
...
  401228:       8b 55 cc                mov    -0x34(%rbp),%edx
  40122b:       8b 45 d0                mov    -0x30(%rbp),%eax
  40122e:       29 c2                   sub    %eax,%edx
  401230:       89 d0                   mov    %edx,%eax
  401232:       89 45 d4                mov    %eax,-0x2c(%rbp)
...
but still 4 line number entries:
...
step-reverse.c           78            0x401228               x
step-reverse.c           78            0x40122b               x
step-reverse.c           78            0x40122e               x
step-reverse.c           78            0x401232               x
...
Note that the insn at 0x401230 doesn't have it's own line number entry.

So, we are at line 79:
...
(gdb) stepi^M
79         callee();    /* STEPI TEST */^M
...
and do a stepi, which lands us, as expected at insn 0x401232:
...
stepi^M
78         a[5] = a[3] - a[4]; /* FINISH TEST */^M
(gdb) PASS: gdb.reverse/step-reverse.exp: simple reverse stepi
print /x $pc^M
$2 = 0x401232^M
(gdb) PASS: gdb.reverse/step-reverse.exp: print /x $pc
...

Then we do a step, which lands us, unexpectedly, at insn 0x40122e:
...
step^M
78         a[5] = a[3] - a[4]; /* FINISH TEST */^M
(gdb) FAIL: gdb.reverse/step-reverse.exp: reverse step into fn call
print /x $pc^M
$3 = 0x40122e^M
(gdb) PASS: gdb.reverse/step-reverse.exp: print /x $pc
...

It's stepping over the 0x401230 insn without line number entry that triggers
the problem.

-- 
You are receiving this mail because:
You are on the CC list for the bug.