[prev in list] [next in list] [prev in thread] [next in thread] 

List:       linux-sparse
Subject:    Re: The stanford checker paper
From:       Christopher Li <sparse () chrisli ! org>
Date:       2004-04-28 16:19:03
Message-ID: 20040428161903.GA21183 () 64m ! dyndns ! org
[Download RAW message or body]

On Wed, Apr 28, 2004 at 12:56:16PM +0100, Dave Jones wrote:
> On Tue, Apr 27, 2004 at 08:22:39AM -0700, Linus Torvalds wrote:
> 
> The database they have at coverity.com is *huge*. They added a bunch of
> extra tests over time vs the old checker, but as you already saw from
> my half-arsed patches last week or so, theres a lot of stuff they pick
> up which is 'noise'. Noise which is still nice to clean up all the same,
> but it's getting harder to seperate the real important nasty bugs from
> the 'trivial' stuff as time goes on.   The 'hundred' reports you mentioned
> above is now in the order of 'thousands'.

It is easy to write checker that catch simple errors and maintain false
positive low. All the bugs in the database already get manual filtered.
It is a labor intensive work.

I can image that the false positive rate will raise if the checker try
to catch more elusive bugs. e.g. touching some member of the structure
without holding certain locks. It needs people familiar with the kernel
to write more complicate checkers.

> What I'm doing with that is every so often, I pick something at random
> (drivers/net for eg) and sift through just those bugs. Even that is a
> quick way to blow away an evening.

I know in some company it is a full time job just to use the meta compiler.

> Ditto. It's a real shame their stuff ended up closed, but at least we

I believe we will get there one day.

Chris
-
To unsubscribe from this list: send the line "unsubscribe linux-sparse" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
[prev in list] [next in list] [prev in thread] [next in thread]