2008/403 [libc printf behaviour for NULL string]

[Darren Reed]

Darren J Moffat wrote:

> Let me set the record straight on the motivation for this case.
>
> It is NOT about "Linux familiarity" even though the GNU libc is used 
> as part of the rationale.  I also used BSD and others pointed out the 
> similar (but not identical) behaviour on AIX and HP-UX.   Basically we 
> are different from everyone else (not Linux but everyone else that 
> matters) and it hurts us.  So please don't put this case into the 
> "Linux familiarity" bucket that isn't why I'm doing this.
>
> This case is about returning to the original SunOS 4.x/BSD behaviour 
> for printf(3C) family.  The reason for doing so is the huge volume of 
> software out there that we want to run on OpenSolaris based 
> distributions that assumes that behaviour.
>
> It might seem like I'm nitpikcing but I've been thinking about running 
> this case for about 8 years now, long before OpenSolaris became real 
> and before the current big "familiarity" push.


It would seem that "%s" can result in "<null string>" inside
the kernel with cmn_err(), at the very least, so libc printing
something is not inconceivable....

See:
$SRC/common/util/string.c

Darren