program test_list real x , y x = 0.0 y = 0.0 write(*,*,iostat=ier) x /y x = 1.0 write(*,*,iostat=ier) x /y end program test_list correctly gave with gfortran 20041010 $ ./a NaN +Infinity but now gives, for either NaN or Infinity, with gfortran 20050108 $ ./a At line 5 of file /cygdrive/d/gfortran/test_list.f90 Internal Error: printf is broken
This works for me on powerpc-darwin with a compiler built yesterday. What target is this?
(In reply to comment #1) > This works for me on powerpc-darwin with a compiler built yesterday. > What target is this? Bother! Sorry, It's on Cygwin/Athlon
*** Bug 19691 has been marked as a duplicate of this bug. ***
Subject: Bug 19363 CVSROOT: /cvs/gcc Module name: gcc Changes by: steven@gcc.gnu.org 2005-02-01 09:06:24 Modified files: libgfortran : ChangeLog libgfortran.h Log message: 2005-02-01 Paul Thomas <paulthomas2@wanadoo.fr> PR libgfortran/19363 PR libgfortran/19691 * libgfortran.h (isfinite): Work around broken isfinite(x) in Cygwin. Patches: http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/libgfortran/ChangeLog.diff?cvsroot=gcc&r1=1.156&r2=1.157 http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/libgfortran/libgfortran.h.diff?cvsroot=gcc&r1=1.22&r2=1.23
Fixeth now, sirrrrrrs?
See http://sources.redhat.com/ml/newlib/2005/msg00099.html for related discussion. This may allow a fix using fixinclude to patch <math.h>.
The patch improves the cygwin test results === gfortran tests === http://gcc.gnu.org/ml/gcc-testresults/2005-02/msg00094.html XPASS: gfortran.dg/ret_pointer_1.f90 execution test XPASS: gfortran.dg/ret_pointer_1.f90 execution test XPASS: gfortran.dg/ret_pointer_1.f90 execution test XPASS: gfortran.dg/ret_pointer_1.f90 execution test XPASS: gfortran.dg/ret_pointer_1.f90 execution test FAIL: gfortran.fortran-torture/execute/forall_3.f90 execution, -O1 === gfortran Summary === # of expected passes 5235 # of unexpected failures 1 # of unexpected successes 5 # of expected failures 21 /usr/local/obj/gcc/gcc/testsuite/../gfortran version 4.0.0 20050202 (experimental)
(In reply to comment #6) > See http://sources.redhat.com/ml/newlib/2005/msg00099.html for related > discussion. This may allow a fix using fixinclude to patch <math.h>. Yes note this is not a gcc problem, the preprocessed source will have: typeof(__x) __x = __x; I will write into the newlib saying that. Closing as fixed as the testsuite is better