https://gcc.gnu.org/bugzilla/show_bug.cgi?id=88342
kargl at gcc dot gnu.org changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|WAITING |RESOLVED
Resolution|--- |FIXED
Assignee|unassigned at gcc dot gnu.org |kargl at gcc dot gnu.org
Target Milestone|--- |9.0
--- Comment #6 from kargl at gcc dot gnu.org ---
Fixed on trunk (aka 9.0). I have implemented the suggested patch in comment #3
for all available REAL kind types for qNaN, sNaN, +inf, and -inf. Thus, for
the code in comment #1, one can now use -ffpe-trap=invalid without getting an
SIGFPE.
Additional work needs to be done where IEEE_VALUE is actually inlined and the
mechanism in the C/C++ frontend are used to get an actual sNaN. That is for
another day.
