On Mon, 13 Jan 2020, Brendan Higgins wrote:
> +KUnit Development
> +open list:KERNEL SELFTEST FRAMEWORK
>
> On Mon, Jan 13, 2020 at 3:40 PM Stephen Rothwell <sfr@xxxxxxxxxxxxxxxx> wrote:
> >
> > Hi Randy,
> >
> > On Mon, 13 Jan 2020 14:59:54 -0800 Randy Dunlap <rdunlap@xxxxxxxxxxxxx> wrote:
> > >
> > > on i386:
> > >
> > > WARNING: modpost: missing MODULE_LICENSE() in drivers/base/test/property-entry-test.o
> > > see include/linux/module.h for more information
> >
> > Sorry, I missed that yesterday.
> >
> > Caused by commit
> >
> > c032ace71c29 ("software node: add basic tests for property entries")
> >
> > from the pm tree interacting with commit
> >
> > 9fe124bf1b77 ("kunit: allow kunit to be loaded as a module")
> >
> > from the kunit-next tree.
>
> Yes, the problem seems to be that the property-entry-test is turned on
> when CONFIG_KUNIT is y or m.
>
> From drivers/base/test/Makefile:
> ...
> obj-$(CONFIG_KUNIT) += property-entry-test.o
>
> It seems we can fix it just by adding the missing MODULE_LICENSE, but
> I think there is a bigger question of whether we should let people do
> this. Do we want to just let people have their tests run whenever
> CONFIG_KUNIT is enabled? I am inclined to think no. It should be
> possible for people to run their test and their test only.
>
I agree completely, CONFIG_KUNIT is too big a switch I
think, aside from the merge issues caused here. I've posted
a patch which introduces a per-test-suite CONFIG option:
https://lore.kernel.org/lkml/1579018183-14879-1-git-send-email-alan.maguire@xxxxxxxxxx/T/#u
Brendan, do let me know if you want me to amend it to add a Suggested-by
from you (didn't want to add it without your permission). Thanks!
Alan
