using integers as booleans

[jayb]

The legacy application that I am porting from g77 to Intel Fortran has many tests of the form:

IF (INTEGER_VAR)

and, since the FORTRAN is generated by RATFOR, many more tests of the form

IF (.NOT.(INTEGER_VAR)) CONTINUE . . .

I would like to detect these with a compiler warning, but even '-warn all' does not flag these.  Is there a flag that I can use?

Side note:  g77 detects these by default, but not when the -fugly flag is used.  And when you use that flag, it looks like some of the tests are executed incorrectly.

Thanks,
Jay

[mecej4]

Use the -stand option. With this option, attempting to compile

subroutine tst(l,m)
integer, intent(in) :: l
integer, intent(out) :: m
m = .not. l
if (m) then
  write(*,*)' m is true'
else
   write(*,*)' m is false'
endif
return
end subroutine tst

gives the warning messages

ilog.f90(4): warning #6188: Fortran 2003 requires a LOGICAL data type in this context.   
m = .not. l
----------^

ilog.f90(5): warning #6188: Fortran 2003 requires a LOGICAL data type in this context.   
if (m) then
----^

[jayb]

Fantastic!  Thanks much.

I'm almost frightened at the prospect of finding all the non-standard features . . . :--)

[TimP]

Don't blame ratfor for this.  I've never seen that extension used with ratfor.  By the way, gfortran should be more reliable than g77 on many current platforms.  It may prove a useful intermediate step.

[jayb]

Howdy.  No blame here.  I meant that since RATFOR translates something like:

if (integer_var) {
   do this }
rest of program

into something like:

IF (.NOT.(INTEGER_VAR)) GO TO  label
DO THIS
label: rest of program

as do most compilers that generate assembly language, or translators that generate C, this feature manifests itself with ".NOT" more often that with the "not"-less expression.

Regarding gfortran, it does not have the options that would be required to compile a working program from the FORTRAN 66/77 hybrid that the legacy program employs.  In a sense, Intel Fortran might theoretically be a stepping stone to an eventual gfortran implementation, as it supports the full range of FORTRAN/Fortran dialects.

JayB