shift count negative or too big, undefined behavior

[mop65715]

Given:

typedef unsigned long long  ulong_long_type ;
istream& istream::operator>>(
   ulong_long_type&        dest )
{
    BGetter                    source( *this ) ;
    ulong_long_type        tmp = source.get() << 56 ;
    tmp |= source.get() << 48 ;
    tmp |= source.get() << 40 ;
    tmp |= source.get() << 32 ;
    tmp |= source.get() << 24 ;
    tmp |= source.get() << 16 ;
    tmp |= source.get() <<  8 ;
    tmp |= source.get()       ;
    return *this ;
}

During compilation I get warnings from Visual studio 2007 surrounding the left shift values 56, 48, 40 and 32, more specifically:

3>c:\sw_dev\msvc_2007-orcas\test\istream.h(158) : warning C4293: '<<' : shift count negative or too big, undefined behavior
3>c:\sw_dev\msvc_2007-orcas\test\istream.h(159) : warning C4293: '<<' : shift count negative or too big, undefined behavior
3>c:\sw_dev\msvc_2007-orcas\test\istream.h(160) : warning C4293: '<<' : shift count negative or too big, undefined behavior
3>c:\sw_dev\msvc_2007-orcas\test\istream.h(161) : warning C4293: '<<' : shift count negative or too big, undefined behavior

Of course I could ignore the warning but the claim of 'undefined behavior' is cause for concern. Thoughts.

Thanks in advance

[Russco]

I dont think msvc understands long long, isn't that a gcc extension?
I think what its done is given you a 32 bit long when you wanted a 64 bit long. Fixable using __int64 but thats microsoft specific. Not 100% sure there is a portable way of declaring a 64 bit int.

[Peter_APIIT]

What is portable for 64bit int ?

Thanks.

[laserlight]

What is portable for 64bit int ?

For C++ at the moment, nothing from the standard, though long long should be available in the next version of the C++ standard, and will probably be guaranteed to be at least 64 bits.

As such, you should use __int64 for Microsoft compilers.

[Russco]

This is one possibility.

[Kheun]

In C99, it also supports int64_t for 64-bit int. So you may like to create a temporary typedef for int64_t so that when Visual C++ finally supports it, you will not have to change much of your code.

typedef __int64 int64_t;

[Lindley]

VS 2005 *does* support long long. That's not the problem. The problem is likely that source.get() is not returning a long long-----therefore, the shift operation is done on a 32-bit value. The program doesn't know or care that the result will be put into a long long after the shift is done.

Try this:

    tmp |= static_cast<ulong_long_type>(source.get()) << 48 ;