Subject: -Wuninitalized comments [was Re: CVS commit: src/sys/dist/pf/net]
To: None <>
From: Luke Mewburn <>
List: tech-userlevel
Date: 06/15/2005 08:27:41
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
Content-Transfer-Encoding: quoted-printable

On Tue, Jun 14, 2005 at 03:51:58PM -0500, James Chacon wrote:
  | On Tue, Jun 14, 2005 at 11:40:59AM +1000, Luke Mewburn wrote:
  | > On Mon, Jun 13, 2005 at 08:33:54PM +0000, James Chacon wrote:
  | >   |=20
  | >   | Module Name:	src
  | >   | Committed By:	jmc
  | >   | Date:		Mon Jun 13 20:33:54 UTC 2005
  | >   |=20
  | >   | Modified Files:
  | >   | 	src/sys/dist/pf/net: pf.c
  | >   |=20
  | >   | Log Message:
  | >   | Fix unitialized warnings that only crop up on m68k. XXGCC taggedd
  | >=20
  | > For consistency with other occurrences of XXXGCC in the source,
  | > could you move these to to separate initializations (after the
  | > variable declaration block) and make the comment text more
  | > consistent (e.g., "XXXGCC -Wuninitialized m68k")?
  | Will do. Are we marking all -Wuninitialized with XXXGCC or only ones th=
  | showed up on specific archs only?

All that we add in 3RDARTY code (**/dist/*), not just MD.

In NetBSD code in some cases I haven't marked the initialization
if it was an initialization that I would have done anyway as a
coding style issue, I may not have done this.

However, if in doubt I used:
	somevar =3D 0;	/* XXXGCC -Wuninitialized */
(after ensuring that `0' was the appropriate initialization value;
sometimes another value such as a code-specific error code was
more appropriate).

(There's an argument to be had that I should have marked more
of them this way).

I also don't just initialize the variable to appease the warning,
because that can mask a bug; careful analysis of the code flow
is required.  I discovered at least 30 bugs this way, including
some potential data corruption issue.


Content-Type: application/pgp-signature
Content-Disposition: inline

Version: GnuPG v1.4.1 (NetBSD)