Linux QA tests (was Re: SMP Linux-37-5)

Ron Holt (ron@caldera.com)
Tue, 20 May 1997 12:51:21 -0600 (MDT)


"Linus Torvalds writes:"
>
> I have found a reason for some occasional hangs (this reason is hopefully
> _the_ reason, but we'll see). My rw-lock implementation had a misfeature
> wrt local interrupts (the rwlock implementation did actually work
> correctly, but required local interrupts to be disabled for readlocks, and
> that was a misfeature).
>
> I'll have a 2.1.37-6 out really soon now with this fixed, once I make it
> go through my torture-tests.

Perhaps this has been discussed before, but is there any interest in
coordinating the maintenance of a set of Linux Quality Assurance test suites?

Some tests that come to mind are:
* The torture tests Linus mentions
* Posix Conformant Test Suite (PCTS) (at least the syscall parts)
* crashme
* Various benchmarks
* Other specialized tests kernel folks have already developed

The ad hoc kernel testing that gets done is great, but having a formal
collection of tests would have several advantages:
* If they were made easy to use, more people would use them
* Bugs could be found more quickly on more hardware combinations
* Linux's "image" in the corporate world would improve (IMHO)
* The kernel would become even better than it is

Caldera would be willing to make available FTP and WWW space for such a
project.

Comments?

Ron

--
Ron Holt <ron@caldera.com> Caldera, Inc.