https://gcc.gnu.org/bugzilla/show_bug.cgi?id=116984
Jakub Jelinek <jakub at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Resolution|--- |INVALID
Status|UNCONFIRMED |RESOLVED
--- Comment #7 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
And note that ubsan instrumentation of array accesses isn't considered a
side-effect for __builtin_object_size/__builtin_dynamic_object_size evaluation
(otherwise the builtins with -fsanitize=undefined would be pretty useless, when
cases like
int index = 0;
__builtin_object_size (&array[index], 0);
would be considered side-effects for which the builtin is documented to return
the "don't know" return value, (size_t)-1 for 0/1, 0 for 2/3 modes).
So, &array[index] isn't considered to have side-effects even when instrumented
and UB in it will be diagnosed (if any).
