linux-kernel - Re: RFC - kernel test result specification (KTAP)

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite for Android: free password hash cracker in your pocket

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Wed, 1 Sep 2021 13:01:55 +0100
From:   Mark Brown <broonie@...nel.org>
To:     kernelci@...ups.io, Tim.Bird@...y.com
Cc:     keescook@...omium.org, rmoar@...gle.com, brendanhiggins@...gle.com,
        davidgow@...gle.com, shuah@...nel.org, dlatypov@...gle.com,
        kunit-dev@...glegroups.com, linux-kselftest@...r.kernel.org,
        linux-kernel@...r.kernel.org, guillaume.tucker@...labora.com
Subject: Re: RFC - kernel test result specification (KTAP)

On Mon, Aug 30, 2021 at 10:04:24PM +0000, Tim.Bird@...y.com wrote:
> > -----Original Message-----
> > From: Kees Cook <keescook@...omium.org>
> > On Mon, Aug 30, 2021 at 05:48:07PM +0000, Tim.Bird@...y.com wrote:
> > > From: Kees Cook <keescook@...omium.org>

> > XFAIL capture the condition of the test logic in that a test for
> > something arm64-specific on a non-arm64 machine is _not_ a "pass". It
> > will fail. But it's an expected failure. And, as a directive, must come
> > with the evidence bit:

> > not ok 1 PAN catches bad accessg # XFAIL PAN CPU feature is arm64 only

> > As a test author, I want to see this as distinct from failure and
> > success.

> I agree that this case is different from a normal failure.

> In the example you provide of a feature that applies to only specific
> architectures, I would use a SKIP instead.  It's not that the testcase failed.
> It's that it does not even make sense for the device under test.

That's what all the arm64 tests are doing at the minute - they're
skipping tests on systems where the feature being tested isn't
supported.  We've determined that the test can't meaningfully be run so
we're going to skip it.  We also don't use XFAIL for tests that are
checking that an error of some kind is generated, the test is formulated
as "error is reported when we do X" meaning that PASS/FAIL makes sense
directly.

There's a couple of XFAIL uses in other selftests but it's pretty
infrequent.

> > This is what we have now except basically just replacing "# $name" with
> > "test $num $name"

> Actually, I like this idea.  Including the number is a nice touch.
> I think that will help to distinguish such a line from other 'unknown line' output.
> I don't know if I'd leave the description off of the result line, though. 
> A lot of parsers are looking for that.

It's also important that humans are able to understand the output and
the descriptions do help with that.

Download attachment "signature.asc" of type "application/pgp-signature" (489 bytes)