https://gcc.gnu.org/bugzilla/show_bug.cgi?id=122879
--- Comment #5 from Fred J. Tydeman <tydeman at tybor dot com> --- In looking at WG14 N3219 (C23), F.2.2, paragraph 6 Recommended practice Any floating-point operator or <math.h> function or macro with a signaling NaN input, unless explicitly specified otherwise, raises an "invalid" floating-point exception. For the operators, there is no qualification as to 'real' operands, so it should apply to 'complex' operands. Besides that Recommended practice (a should), there is in paragraph 3, it is implementation defined. So, the current gcc behaviour does not violate C23. It could be made more consistent for a slight slowdown in code speed.
