right/left shifts with (width of type)

[Nathan Sidwell]

Stefan Meixner wrote:
> 
> Hi,
> 
> right now I will not call it a bug, but gcc/egcs
> practise with right/left shifts with counts equal
> to the width of the type is at least not protable
> (at least on a i486-pc-linux-gnulibc1 where I tested it).
> 
> I would expect the following expression to hold
> for both << and >> operators
> 
> unsigned long u = ~0;
> 
>     if ( (u << (sizeof(unsigned long)*CHAR_BIT)) == 0 )
>           printf("GOOD");
>     else
>           printf("BAD");
> 
> At least this is the behavior I would expect and I get from
> the other compilers I tested.
> 
> I know that this maybe implementation defined. But shifting
> any unsigned long by count=width, the result should be zero.
> 
> Am I wrong?
You are wrong. In both C and C++ it is undefined if the shift amount is
negative or greater or equal to the length in bits of the promoted left
operand. That's *undefined*, not implementation defined, anything can
happen.

Your assertion that x << bits_in_x == 0 is not true. Relying on it is
inherently non-portable -- it's not even guaranteed to be consistent
within a single program.

nathan
-- 
Dr Nathan Sidwell :: Computer Science Department :: Bristol University
      You can up the bandwidth, but you can't up the speed of light      
nathan@acm.org  http://www.cs.bris.ac.uk/~nathan/  nathan@cs.bris.ac.uk