[nycbug-talk] BSD Chapter in HLE

[Ray Lai]

On Fri, Sep 15, 2006 at 06:24:28PM -0400, Dru wrote:
> On Fri, 15 Sep 2006, Ray Lai wrote:
> >>What about Coverity
> >>audits being integrated into engineering processes?
> >
> >Coverity is a nice tool, but its suggested fixes should not be committed
> >wholesale without checking if they are correct.  This is true for just
> >about every other tool.  Don't overlook lint, either.  Chad Loder has
> >been improving our lint to quiet it down and to concentrate on real
> >issues.  It is pretty useful to run these tools on the source code and
> >look carefully at areas they point out, concentrating on new findings.
> >Be careful not to change code just to silence the tools, however; this
> >can introduce bugs or silence legitimate ones.
> 
> I'd like to stress the quality of code and the release engineering, commit 
> bit processes as this is a big difference between the BSDs and Linux. I'm 
> also not a committer so it would be interesting to have a paragraph or so 
> from each project explaining how their processes promote secure and 
> quality code.

Here is my interpretation of the OpenBSD process:

We generally require other developers to okay commits, so obvious bugs
and other questionable changes don't creep in.  The tree must never
be broken.  It's better to make small, verifiable changes to achieve
a larger goal than to make huge, difficult to understand commits.
Manual or tool-aided audits should be performed occasionally; I like
doing these on old code because there are usually obvious bugs to fix
and because I am unfamiliar with the code, I am forced to learn what it
does (and not what the author intended it to do).  It's also important
to do this on new code, of course.

-Ray-