Re: [PATCH] New option to support asymmetric integer ranges in gfortran

[Scott Robert Ladd <coyote at coyotegulch dot com>]

Steve Kargl wrote:
> I can't check the ISO C standard, but I think appealing to the
> interoperability of Fortran and C is not a very good argument.

From 7.20.6.1 of ISO/IEC 9899:1999 ("Standard C"):

The abs, labs, and llabs functions compute the absolute value of an 
integer j. If the result cannot be represented, the behavior is 
undefined. 256)

Footnote:
256) The absolute value of the most negative number cannot be 
represented in two's complement.

In other words, we're consistent in that the absolute value of the most 
negative integer is undefined in the standards. As such, my proposed 
patch allows gfortran to behave in conformance with standard practice.

As for interoperability: It is one of the foci in the proposed Fortran 
"2003" Standard. The ISO_C_BINDING intrinsic module defines kinds that 
will to operate across the Fortran/C boundary... C_INT32_T, for example.

Perhaps, for Fortran 2003, we would define a kind of C_INT32_T as 
different from a kind of 4. That way, the native integer(kind=4) could 
impose Fortran's integer model (which rejects -2147483648), while 
integer(kind=C_INT32_T) allows -2147483648. In fact, this may be the 
very intent of the new standard.

Just a thought.

--
Scott Robert Ladd
Coyote Gulch Productions (http://www.coyotegulch.com)
Software Invention for High-Performance Computing