https://gcc.gnu.org/bugzilla/show_bug.cgi?id=88074
Richard Biener <rguenth at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Attachment #45039|0 |1
is obsolete| |
--- Comment #21 from Richard Biener <rguenth at gcc dot gnu.org> ---
Created attachment 45052
--> https://gcc.gnu.org/bugzilla/attachment.cgi?id=45052&action=edit
patch
For reference this is what I tested. It FAILs the following at all
optimization
levels
FAIL: gfortran.dg/norm2_3.f90 -O0 (test for excess errors)
UNRESOLVED: gfortran.dg/norm2_3.f90 -O0 compilation failed to produce
executa
ble
/space/rguenther/src/svn/trunk/gcc/testsuite/gfortran.dg/norm2_3.f90:24:9:
Error: Arithmetic overflow at (1)^M
/space/rguenther/src/svn/trunk/gcc/testsuite/gfortran.dg/norm2_3.f90:24:9:
Error: Result of ABS overflows its kind at (1)^M
compiler exited with status 1
that is
if (abs (NORM2 ([real(qp) :: 1, 2, huge(3.0_qp)]) - huge(3.0_qp)) &
> epsilon(0.0_qp)*huge(3.0_qp)) STOP 1
with
integer,parameter :: qp = selected_real_kind (precision (0.0d0)+1)
