[Bug preprocessor/55715] New: bogus overflow warning for #if A-B when A<0 & B==minimum integer

[eggert at gnu dot org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55715



             Bug #: 55715

           Summary: bogus overflow warning for #if A-B when A<0 &

                    B==minimum integer

    Classification: Unclassified

           Product: gcc

           Version: 4.7.2

            Status: UNCONFIRMED

          Severity: normal

          Priority: P3

         Component: preprocessor

        AssignedTo: unassig...@gcc.gnu.org

        ReportedBy: egg...@gnu.org





Consider this C program:



#if 9223372036854775807 < 0

#endif

#if -1 - 9223372036854775807 < 0

#endif

#if -1 - (-1 - 9223372036854775807) < 0

#endif



When compiled with GCC (x86-64), GCC 4.7.2 warns:



t.c:5:37: warning: integer overflow in preprocessor expression [enabled by

default]



The warning is incorrect, since there is no overflow in this program.



Evidently GCC's preprocessor computes A-B by first negating B, and then adding

the result to A.  This causes GCC to incorrectly report overflow when B is the

minimum possible integer and A is negative, because in this case -B overflows

even though A-B does not overflow.