http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55462
Bug #: 55462
Summary: [4.8 Regression] FAIL: gfortran.dg/quad_2.f90 -O*
execution test after revision 193770
Classification: Unclassified
Product: gcc
Version: 4.8.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: libquadmath
AssignedTo: unassig...@gcc.gnu.org
ReportedBy: domi...@lps.ens.fr
CC: bur...@gcc.gnu.org, h...@gcc.gnu.org
After revision 193770, the test gfortran.dg/quad_2.f90 fails at runtime. As
shown by the following reduced test, this is due to a difference between the
last ULP for read and write.
program test_qp
use iso_fortran_env, only: real_kinds
implicit none
integer, parameter :: QP = real_kinds(ubound(real_kinds,dim=1))
real(qp) :: fp2, fp4
character(len=80) :: str4
fp2 = sqrt (2.0_qp)
write (str4,'(g0)') fp2
read (str4, *) fp4
print *, fp2, fp4, (fp4-fp2)/epsilon(fp2)
if (fp2 /= fp4) call abort()
end program test_qp
The output is
1.41421356237309504880168872420969798
1.41421356237309504880168872420969779
-1.00000000000000000000000000000000000
