Am 29.03.2015 um 17:36 schrieb Hajime Tazaki:

At Sat, 28 Mar 2015 22:17:40 +0100,
Richard Weinberger wrote:

quoted

quoted

Continuous testing is paramount.  Running the kernel as
a lib provides an unparalleled method for testing most of
the kernel.  It will improve testing capabilities
dramatically,
and on the flipside it will keep the libos working.
Everyone wins.

If it can be done cheap, yes. But our in-kernel tests improved over the years a lot.
Now have lockdep, KASan, kmemleak, etc. to find *real-world* issues and the need for stubbed testing
decreases.

let me take the same example I raised.

- Patchwork [net-next] xfrm6: Fix a offset value for network header in _decode_session6
http://patchwork.ozlabs.org/patch/436351/

without stubbed testing (I didn't know this term btw), we
can't decrease untested paths of the code.

the above bug is for Mobile IPv6, which not so many people
are using though, but it's certainly a regression for a
person.

testing framework with libos is based on a network
simulator, with a slight decreased realism (but it can
detect a real bug !), but provides a lightweight multi-node
testing framework with a single test scenario script to
control over the nodes.

it doesn't require heavyweight machines nor complex cabling
for a bunch of tests.

even a framework is not cheap, I would use such a testing
tool IF we can improve the code. plus (as you may know), it
certainly reduces the maintenance effort once it's automated.

Don't get me wrong, I totally agree that this kind of testing is good.
But as I said before we have to keep the maintenance burden in mind.

Let's wait a bit what Arnd says. He is the Linux arch maintainer.

Thanks,
//richard