Double precision loss on platforms with 64-bit integers on expression having ints exclusively

[alcz]

Does not seem to be C compiler specific (other than having 64-bit ints), plus it's not Clipper compatible number scope to rely on bignum arithmetics in portable code.

  LOCAL i
   
   // intro
   ? i := - hb_HexToNum( hb_StrToHex( "ABCDEFG" ) )
   ? hb_HexToStr( hb_NumToHex( Abs( i ) ) ) // -> ABCDEFG

   ? i := hb_HexToNum( hb_StrToHex( "ABCDEFG" ) ) * -1
   ? hb_HexToStr( hb_NumToHex( Abs( i ) ) ) // -> ABCDEFH (surprise)

   // plain tests
   i := 18368730161825351 
   ? i * -1        // -18368730161825352
   ? Int( i ) * -1 // -18368730161825352
   ? - i           // -18368730161825351

   // with precompiled constants does not happen
   ? 18368730161825351 * -1 // -18368730161825351
   ? - 18368730161825351    // -18368730161825351

https://os.allcom.pl/hb/#!eJydUttOwkAQfecrTnhqTRosCBITNQgqJkQf5N20ZaGbtLvNzhbx753tBYmIMW7TtDM5t80MgMXLdLKA7ABwb6-HmS7jTKAwIpEktUKmicDfIovsWpuc8C5titFFEEsLqazYCFMhxI5ZVJG4TVwnWUlyK7KPRnylBUFpCxIih9WIBaZIdF7ITBhQwaZrmcDTNuXappFCGm2l2hz4kd9p5Lgw2v3fQuLqGgHS-G0udkv9XOaeK16tWWrueOhO7qaz-4fHLnz4NacFM6gCM6kBT2LyWNKvHjYKbtDQOwd2fzTDGYLw345zeFSawkgS-3vzKKSCFWTJdaow4XgwGl8OzsNROO4PB8MQbVJnj-Y45e_Ifg18UrZOUOF_AQaM-jo_AcM2aLUpbpXqCa941Ips5JZjvwppVBRC1dLHlzgZJmzDHHNOpvoESy7E6w