Build-time tests?

Build-time tests?

[Joe Lawrence]

Hello kbuild, kselftest,

I've been working on a patchset which adds an additional build script to
the toolchain when compiling livepatches.  There are a few kernel
section features in which this script does not yet support, but can
detect and abort when it encounters.  To test this detection, I've
written a small set of kernel modules that require such sections.

A few questions: 

Is build-testing out of scope for kernel selftests?  For expediency,
it was really easy to spin out new lib/livepatch kernel modules.

Does kbuild support the notion of expected failure?  In this case, the
build script returns a non-zero error and the build stops.

Am I trying to fit a square peg in a round hole?  I could easily keep
these build tests in a private branch, but could they exist in a
different format somewhere else in the tree?

Suggestions welcome,

-- Joe

Re: Build-time tests?

[Masahiro Yamada]

On Thu, May 9, 2019 at 4:30 AM Joe Lawrence <joe.lawrence@redhat.com> wrote:
>
> Hello kbuild, kselftest,
>
> I've been working on a patchset which adds an additional build script to
> the toolchain when compiling livepatches.  There are a few kernel
> section features in which this script does not yet support, but can
> detect and abort when it encounters.  To test this detection, I've
> written a small set of kernel modules that require such sections.
>
> A few questions:
>
> Is build-testing out of scope for kernel selftests?  For expediency,
> it was really easy to spin out new lib/livepatch kernel modules.
>
> Does kbuild support the notion of expected failure?  In this case, the
> build script returns a non-zero error and the build stops.


If I understood correctly, your are asking about kselftest,
maintained by Shuah Khan.

(Unfortunately), makefiles under tools/ adopted completely
different build system, and it is out of scope of kbuild.


> Am I trying to fit a square peg in a round hole?  I could easily keep
> these build tests in a private branch, but could they exist in a
> different format somewhere else in the tree?
>
> Suggestions welcome,
>
> -- Joe



-- 
Best Regards
Masahiro Yamada