https://gcc.gnu.org/bugzilla/show_bug.cgi?id=118074
--- Comment #4 from Andrew Pinski <pinskia at gcc dot gnu.org> --- (In reply to Weibo He from comment #3) > Intert a printf into final_suspend helps clang and GCC 10 ~ 14 give expected > result. While GCC trunk not work. > Is it a GCC 15 regression? It does not help -O0 gcc nor -O0 clang, both still give runtime errors with -fsanitize=address . clang -O2 does not give a runtime error with -fsanitize=address but I can only assume it is able to optimize away the load after the delete . So I tried MSVC with /fsanitize=address but the dll is not installed on godbolt so I have no way to fully test it.
