[Bug fortran/38823] Diagnose and treat (-2.0)**2.0 properly

[burnus at gcc dot gnu dot org]

------- Comment #12 from burnus at gcc dot gnu dot org  2009-01-13 21:41 -------
>  Add a runtime check that x in x**y is not < 0

I'm actually against a run time check which is based on -std=, -pedantic, -W*
or similar compile time diagnostic flags. I regard such trickery behind the
scenes as evil and I would expect such an error only for -fcheck*

Note, there is already a check for
  x = -1.9
  y = (-2.0)**(x)
It gives NaN - and with trapping that is a SIGFPE. I think that should good
enough, but I don't mind having a -fcheck* option adding a run-time check.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38823