Re: CVS commit: src/lib/libc/string

[Joerg Sonnenberger <joerg%britannica.bec.de@localhost>]

On Fri, Nov 25, 2011 at 04:23:46PM +0000, David Holland wrote:
> On Fri, Nov 25, 2011 at 02:48:30PM +0100, Joerg Sonnenberger wrote:
>  > > > > > > Sure it does. You should just have written the code using 
> (size_t)1,
>  > > > > > > or even just 1UL, instead of 1ULL. There is no port where size_t 
> is
>  > > > > > > unsigned long long.
>  > > > > > > 
>  > > > > > > ... Unless what you meant was "get rid of all 32-bit ports" :-)
>  > > > > > 
>  > > > > > Well, there is Win64 and this is essentially portable code...
>  > > > > 
>  > > > > so use (size_t)1 then.
>  > > > 
>  > > > Point remains that the original code is correct and no overflow can
>  > > > happen. As such lint's behavior is just bogus.
>  > > 
>  > > Warning about 64->32 truncation is useful. Coding to avoid unnecessary
>  > > ones is straightforward. I think you're complaining only because you
>  > > got caught out by it :-)
>  > 
>  > Which part of "it can't truncate" is hard to understand? I am
>  > complaining because it is no reason to error out.
> 
> And why should the/any static analyzer have to figure that out when
> coding it right makes it a nonissue?

Because that's what is making warnings useful. To not complain about
non-issues.

Joerg