https://gcc.gnu.org/bugzilla/show_bug.cgi?id=63914
--- Comment #3 from Richard Biener <rguenth at gcc dot gnu.org> ---
We try to transition from
{ Inf, 0.0, 0.0, 0.0 } * { 0.0, 0.0, 0.0, 0.0 } == { Nan, 0.0, 0.0, 0.0 }
to
VARYING * { 0.0, 0.0, 0.0, 0.0 } == { 0.0, 0.0, 0.0, 0.0 }
which is... "interesting". -ffast-math allows for this as it considers
-ffinite-math-only for which the first constant folding returns an
"invalid" number. Hmm.
We have canonicalize_value that is supposed to catch this but it fails to
handle VECTOR_CSTs or COMPLEX_CSTs properly.
