https://gcc.gnu.org/bugzilla/show_bug.cgi?id=90017
Richard Biener <rguenth at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Keywords| |wrong-debug
CC| |aoliva at gcc dot gnu.org,
| |jakub at gcc dot gnu.org
--- Comment #1 from Richard Biener <rguenth at gcc dot gnu.org> ---
Hmm, I think the debuginfo is "correct", at the last invocation of
optimize_me_not () l is indeed 8. At -O3 we "merely" unrolled the
inner loop completely. When you step through the program at
each optimize_me_not invoation the value of l is correct but gdb
seems to set only one breakpoint for 'b 15' which you can see doing
(gdb) b optimize_me_not
(gdb) b 15
(gdb) run
Breakpoint 1, optimize_me_not () at t.c:2
2 __asm__ volatile ("" : : : "memory");
(gdb) c
Continuing.
Breakpoint 1, optimize_me_not () at t.c:2
2 __asm__ volatile ("" : : : "memory");
(gdb) c
Continuing.
Breakpoint 1, optimize_me_not () at t.c:2
2 __asm__ volatile ("" : : : "memory");
(gdb) c
Continuing.
Breakpoint 1, optimize_me_not () at t.c:2
2 __asm__ volatile ("" : : : "memory");
(gdb) c
Continuing.
Breakpoint 2, main () at t.c:15
15 optimize_me_not();
looking at disassembly with source interleaved shows that this might get
wrong somewhere during debug-info creation, not sure exactly how we
compute the line number program. Maybe it's also a consumer issue.
