Re: [PATCH v5 3/6] thunderbolt: test: use NULL macros

[Mika Westerberg <mika.westerberg@linux.intel.com>]

On Fri, Feb 11, 2022 at 06:26:56PM +0200, Mika Westerberg wrote:
> > To test it I had enabled:
> > PCI, USB4 and USB4_KUNIT_TEST
> > 
> > and then run it with
> > 
> > ./tools/testing/kunit/kunit.py run --jobs=$(nproc) --arch=x86_64
> > 
> > Unfortunately, kunit was not able to run the tests
> > 
> > This hack did the trick:
> > 
> > 
> >  int tb_test_init(void)
> >  {
> > -       return __kunit_test_suites_init(tb_test_suites);
> > +       //return __kunit_test_suites_init(tb_test_suites);
> > +       return 0;
> >  }
> > 
> >  void tb_test_exit(void)
> >  {
> > -       return __kunit_test_suites_exit(tb_test_suites);
> > +       //return __kunit_test_suites_exit(tb_test_suites);
> >  }
> > +
> > +kunit_test_suites(&tb_test_suite);
> > 
> > I looked into why we do this and I found:
> > 
> > thunderbolt: Allow KUnit tests to be built also when CONFIG_USB4=m
> > 
> > 
> > I am a bit confused. The patch talks about build coverage, but even
> > with that patch reverted if
> > USB4_KUNIT_TEST=m
> > then test.c is built.
> > 
> > Shouldn't we simply revert that patch?
> 
> Nah, either build it into the kernel or load the driver manually:
> 
>   # modprobe thunderbolt

Forgot to explain why this does not run the tests (I think):

 ./tools/testing/kunit/kunit.py run --jobs=$(nproc) --arch=x86_64

The driver depends on PCI and I don't think that's enabled on UML at
least. I typically run it inside QEMU.