Re: RFC - kernel selftest result documentation (KTAP)

From: Frank Rowand
Date: Fri Jun 19 2020 - 14:47:35 EST


On 2020-06-16 18:58, Kees Cook wrote:
> On Tue, Jun 16, 2020 at 12:44:28PM -0700, Brendan Higgins wrote:
>> On Tue, Jun 16, 2020 at 9:42 AM Bird, Tim <Tim.Bird@xxxxxxxx> wrote:
>>>> From: Paolo Bonzini <pbonzini@xxxxxxxxxx>
>>>>
>>>> On 15/06/20 21:07, Bird, Tim wrote:
>>>>>> Note: making the plan line required differs from TAP13 and TAP14. I
>>>>>> think it's the right choice, but we should be clear.
>>>>
>>>> As an aside, where is TAP14?
>>> By TAP14, I was referring to the current, undocumented, KUnit
>>> conventions.
>>
>> Not so. TAP14 is the proposed next version of TAP13:
>>
>> https://github.com/TestAnything/testanything.github.io/pull/36
>> https://github.com/isaacs/testanything.github.io/blob/tap14/tap-version-14-specification.md
>
> I was reading this (I haven't compared to the blob above):
>
> https://github.com/TestAnything/Specification/blob/tap-14-specification/specification.md
>
>> Based on the discussion, it seems like most of the things we wanted
>> from TAP14 would probably make it in if TAP ever accepts another pull
>> request.
>
> Were our leading diagnostic lines part of their latest spec? I thought
> we were pretty far off in left field for that particular bit.
>
>>> My personal preference is to have the dash. I think it's more human readable.
>>> I note that the TAP spec has examples of result lines both with and without
>>> the dash, so even the spec is ambiguous on this. I think not mandating it
>>> either way is probably best. For regex parsers, it's easy to ignore with '[-]?'
>>> outside the pattern groups that grab the number and description.
>>
>> I don't think we care, because we don't use it.
>
> Yeah, I'm in the same place. I don't care -- I would just like a
> determination. (The "implied" nature of it in TAP14 bothers me.)
>
>>>> XFAIL/XPASS are different from SKIP. I personally don't have a need for
>>>> them, but kselftests includes XFAIL/XPASS exit codes and they aren't
>>>> reflected into selftests/kselftest/runner.sh.
>>>>
>>>> Likewise, kselftest.h has ksft_inc_xfail_cnt but not
>>>> ksft_test_result_xfail/ksft_test_result_xpass.
>
> I proposed fixing that recently[1]. seccomp uses XFAIL for "I have
> detected you lack the config to test this, so I can't say it's working
> or not, because it only looks like a failure without the config."

Based on that description, the case sounds like it should be a skip.

Or if the entire test depends on the missing config then Bail out might
be appropriate.

>
> [1] https://lore.kernel.org/lkml/20200611224028.3275174-7-keescook@xxxxxxxxxxxx/
>