https://gcc.gnu.org/bugzilla/show_bug.cgi?id=125603
Richard Biener <rguenth at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Last reconfirmed| |2026-06-05
Ever confirmed|0 |1
CC| |hubicka at gcc dot gnu.org
Status|UNCONFIRMED |NEW
--- Comment #1 from Richard Biener <rguenth at gcc dot gnu.org> ---
Confirmed.
#1 0x000000000128a039 in profile_probability::apply_scale (
this=0x7fffffffd134, num=1, den=0) at ../../src/gcc/gcc/profile-count.h:554
554 safe_scale_64bit (m_val, num, den, &tmp);
#3 0x000000000194c858 in set_even_probabilities (
bb=<basic_block 0x7ffff743eea0 (7)>, unlikely_edges=0x7fffffffd1d0,
likely_edges=0x7fffffffd210) at ../../src/gcc/gcc/predict.cc:968
968 e->probability = remainder / count;
(gdb) l
963 remainder -= (profile_probability::very_unlikely ()
964 * unlikely_count);
965 int count = nedges - unlikely_count - 1;
966 gcc_assert (count >= 0);
967
968 e->probability = remainder / count;
I guess the assert should have been count > 0 (which then would have tripped as
well)?
We have a switch with all outgoing edges predicated unlikely ...
