Subject: Re: port-xen/29887: sysctl kern.consdev coredumps
To: Jason Thorpe <thorpej@shagadelic.org>
From: Bill Studenmund <wrstuden@netbsd.org>
List: tech-userlevel
Date: 06/21/2005 16:35:44
--FN+gV9K+162wdwwF
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
Content-Transfer-Encoding: quoted-printable

On Tue, Jun 21, 2005 at 09:54:21AM -0700, Jason Thorpe wrote:
>=20
> On Jun 20, 2005, at 7:33 PM, der Mouse wrote:
>=20
> >Maybe that's because it isn't in the manual page.
>=20
> That's precisely my point.

I don't think it is. If it is, you're making your arguement very=20
indirectly.

Yes, whomever added this behavior to printf(3) should have updated the=20
manual page over a decade ago.

However the arguement (in general) hasn't been, "Oh, if we'd known=20
printf(3) was supposed to do this we would have told you to add option BAR=
=20
to the gcc config." If it had been, we (I at least) would have said, "Yes,=
=20
the printf(3) documentation was incomplete. We are sorry. What do we now=20
do to get that behavior with gcc?"

The arguements have been more, "It's the compiler's choice" or "It is=20
better to core dump." Which aren't related to the documentation.

Given how many times NetBSD has been a "UTSL" project, I don't see how the
fact that man 3 printf doesn't describe the %s NULL behavior really can
justify a decision going forward. Helping explain how we got where we are,=
=20
most certainly! But not justify a course of action.

To be honest, I think we should just adjust puts(). I think the=20
optimization that gcc is performing can be a good one, and I'd rather=20
adjust puts() than lose the optimization. :-)

Take care,

Bill

--FN+gV9K+162wdwwF
Content-Type: application/pgp-signature
Content-Disposition: inline

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.3 (NetBSD)

iD8DBQFCuKRQWz+3JHUci9cRAnmBAJsEPBNJT9sLozNDTqlIr5i2gur8nQCfV7AL
2RzQsR75GBJcGTbYSzj2kj4=
=0gUt
-----END PGP SIGNATURE-----

--FN+gV9K+162wdwwF--