Re: [PATCH 2/2] Kprobes: Move kprobes examples to samples/

[Ananth N Mavinakayanahalli <ananth@in.ibm.com>]

On Thu, Jan 03, 2008 at 10:18:06AM -0500, Mathieu Desnoyers wrote:
> * Ingo Molnar (mingo@elte.hu) wrote:
> > 
> > * Ananth N Mavinakayanahalli <ananth@in.ibm.com> wrote:
> > 
> > > > feature request: please make this work in the !modular case as well 
> > > > - if built-in then it should just run sometime during bootup and run 
> > > > the tests and report success/failure. This way automated testing can 
> > > > pick up any regressions much easier.
> > > 
> > > Will try cook up something along those lines. It'll be easy to verify 
> > > if the probes inserted and removed properly, but verifying handlers 
> > > run correctly will need some work.
> > > 
> > > We have a sort of regression test bucket that uses expect to parse the 
> > > dmesg to verify handlers did run correctly; that isn't a totally 
> > > in-kernel solution anyway. I have a couple of ideas in mind to make it 
> > > easier.
> > 
> > Great. Would be really nice to have something along the lines of 
> > CONFIG_DEBUG_LOCKING_API_SELFTESTS. Those unit tests took time to 
> > develop, but they caught more than 90% (!) of the internal lockdep 
> > engine bugs before they ever hit mainline.
> > 
> 
> I would just like to point out that the samples/ directory should keep
> files as easy to read and understand for newcomers (it is meant to be
> compiled Documentation examples). I see the interest in turning it into
> a regression test too, but I would recommend leaving the "test" code out
> of the sample module itself to improve readability.

Agreed. I am working on a test bucket that doesn't touch the samples.
It'll live on its own, helping with boot time smoke tests.

Ananth