With the code here: ---- program pb character*10 form character*4093 str str=' ' write(form,120) 3,0 write(str(1:3),form) 1.0d0 - 1.110223024625157D-16 print *,str(1:10) 120 format('(f',i2,'.',i2,')') end ---- Compiled with gfortran 4.3.2: the output is 2 With the gfortran 4.2.4 the result is 1 (which is the right one) $ gfortran-4.3 -o pb pb.f && ./pb 2. $ gfortran-4.2 -o pb pb.f && ./pb 1.
Confirmed on trunk, with this test: program pb write(*,'(F3.0)') 1.0d0 - 1.110223024625157D-16 end Changing decimals display to a value larger than 0 outputs 1 correctly.
Jerry, do you have an idea?
I will try to look at this tonight.
Subject: Bug 37863 Author: jvdelisle Date: Sat Oct 18 05:27:23 2008 New Revision: 141208 URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=141208 Log: Fix ChangeLog 2008-10-17 Jerry DeLisle <jvdelisle@gcc.gnu.org PR libfortran/37863 * io/write_float.def (WRITE_FLOAT): Round to 1.0 correctly. * io/io.h (st_parameter_44): Fix id type declaration. Modified: trunk/libgfortran/ChangeLog
Fixed on trunk. I accidentally committed the patch with the wrong PR number in the Changelog (37707) This I will backport to 4.3 in a day.
I must say that I am very impressed by your reactivity! Thanks!
Subject: Bug 37863 Author: jvdelisle Date: Sun Oct 19 15:28:25 2008 New Revision: 141227 URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=141227 Log: 2008-10-19 Jerry DeLisle <jvdelisle@gcc.gnu.org PR libfortran/37863 Backport from trunk. * io/write_float.def (WRITE_FLOAT): Round to 1.0 correctly. 2008-10-19 Jerry DeLisle <jvdelisle@gcc.gnu.org PR libfortran/37707 Backport from trunk. * io/list_read.c (nml_get_obj_data): If the first namelist object rank is greater than zero, call nml_object_read with the first object rather than the sub-object. Modified: branches/gcc-4_3-branch/libgfortran/ChangeLog branches/gcc-4_3-branch/libgfortran/io/list_read.c branches/gcc-4_3-branch/libgfortran/io/write_float.def
Fixed on 4.3 and 4.4