svn commit: trunk/busybox: archival coreutils miscutils modutils ne etc...

[Rich Felker]

On Sat, Dec 30, 2006 at 11:31:17PM +0100, Denis Vlasenko wrote:
> On Saturday 30 December 2006 23:26, Yann E. MORIN wrote:
> > On Saturday 30 December 2006 185, vda at busybox.net wrote:
> > > done a dozen of randconfig test. guess what? ALL failed...
> > > these are resulting fixes
> > 
> > OK, some more results from randconfig:
> >  - mkfs.minix
> >  - less
> >  - ping6
> >  - od (bloaty)
> > 
> > Patches attached.
> 
> Thanks, applying...
> 
> I wonder why on my system gcc doesn't warn about size_t in printf??

Due to legacy stupidity, size_t is unsigned int rather than unsigned
long on 32bit archs. Both would have the same meaning except that
correcting it to use unsigned long would cause warnings for code that
will break on 64bit archs. On all unix archs, sizeof(size_t) and
sizeof(long) are the same, so it makes sense for size_t to always be
defined as unsigned long.

Incidentally, I think the reason it's kept as unsigned int is to allow
legacy code that printf's size_t variables to work without warnings...
Yet another reason the gcc maintainers are incompetent.

Rich