------- Comment #4 from tkoenig at gcc dot gnu dot org 2007-02-16 20:50 ------- (In reply to comment #3)
> There is a new -Wconversion implementation. Perhaps you could do the same for > fortran as I did for C/C++. Take a look at c-common.c (conversion_warnings). > It > would be great if -Wconversion behaves the same way (more or less) in all > front-ends. This would be a good idea. Looking at the code from c-common.c, we should warn about real(kind=4) :: a real(kind=8) :: b integer(kind=1) :: i1 integer(kind=4) :: i4 i4 = 2.3 ! Not exact i1 = 500 ! doesn't fit a = 2**26-1 ! Loses bits of precision b = 1d99 ! Doesn't fit the type a = i4 ! may lose extra digits b = i4 ! This is OK (enough digits in the kind=8 var) i1 = i4 ! i1 may not be able to represent the values in i4 a = b ! precision loss For Fortran, we should also warn about b = 2.5 print *,b**(3/2) ! identical to b**1, usually unintended and maybe even about print *,4/3 ! constant integer division -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=27866
