Subject: Re: netinet compilation warnings with IPv6 code, but without INET6
To: Todd Vierling <tv@pobox.com>
From: Simon Burge <simonb@netbsd.org>
List: current-users
Date: 07/04/1999 00:06:14
Todd Vierling wrote:

> On Fri, 2 Jul 1999, Erik Bertelsen wrote:
> 
> : On my pmax, I observed the kernel problems shown below while compiling
> : a kernel based on yesterday's sources. It does not contain any of the
> : new IPv6 options (e.g. INET6):
> : 
> : k make -k
> : cc  -O2 -Wall -Werror -Wstrict-prototypes -Wmissing-prototypes
> : -Wuninitialized -Wpointer-arith -Wno-main -G 0  -mno-abicalls
>    ^^^^^^^^^^^^^^
> Uh, `whassis?'
> 
> If `cc -v' reports `egcs-1.1*', turn this off.  egcs gets this very, very
> wrong and we have it explicitly disabled from -Wall for a reason.

-Wuninitialized has helped me catch a number of bugs on the pmax - I
can't remember how many were in MI code vs. MD code.  From Jason
earlier in this thread:

> No, most ports don't use -Wuninitialized due to EGCS getting it wrong
> some large amount of the time on the i386 and m68k ports.  but it seems
> to be usually right on the MIPS.

Do we know in this particular case whether or not it was a false
positive?  Until INET6, pmax kernels were building happily for a
couple of months with this option.

FWIW, sparc64 seems to be the only other port that uses -Wuninitialized...

Simon.