[Bug libfortran/96613] SIGFPE on min1() with -ffpe-trap=invalid switch

[kargl at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=96613

kargl at gcc dot gnu.org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2020-08-14
     Ever confirmed|0                           |1
                 CC|                            |kargl at gcc dot gnu.org
           Priority|P3                          |P4

--- Comment #1 from kargl at gcc dot gnu.org ---
Compiling with -fdump-tree-original gives

  integer(kind=4) ix;
  real(kind=4) x;

  x = 7.7643945e+9;
  {
    integer(kind=4) M.0;

    M.0 = 5000;
    M.0 = MIN_EXPR <(integer(kind=4)) x, M.0>;
    ix = M.0;
  }

The MIN_EXPR is clealy a bad idea as x > huge(1).

The Fortran standard states, that for example,
MIN1(a,b) = INT(MIN(a,b)).  I would have expected
to see something along the lines of

  integer(kind=4) ix;
  real(kind=4) x;

  x = 7.7643945e+9;
  {
    real(kind=4) M.0;

    M.0 = 5000.0;
    M.0 = MIN_EXPR <x, M.0>;
    ix = (integer(kind=4)) M.0;
  }

If this is effecting your code, I suggest changing it to
min1() = int(min()) to work around the issue.  This might
get fixed someday when a gfortran contributor has time to
look at it.