On Tue, 2019-12-17 at 11:49 -0500, Stephen Smalley wrote:
> On 12/17/19 11:38 AM, Richard Haines wrote:
> > On Tue, 2019-12-17 at 10:36 -0500, Stephen Smalley wrote:
> > > On 12/16/19 9:09 AM, Stephen Smalley wrote:
> > > > On 12/15/19 12:06 PM, Richard Haines wrote:
> > > > > Test filesystem permissions using mount(2)/umount(2).
> > > > >
> > > > > From kernels 5.5 filesystem { watch } is also tested.
> > > > >
> > > > > Signed-off-by: Richard Haines <
> > > > > richard_c_haines@xxxxxxxxxxxxxx>
> > > >
> > > > This didn't pass travis-ci, looks like a combination of failing
> > > > check-syntax and FAN_MARK_FILESYSTEM not being defined (maybe
> > > > the
> > > > kernel
> > > > headers are too old in the base distro?).
> > >
> > > Possibly we need to install our own kernel headers for the
> > > testsuite?
> > I assume this is on the travis system (that I don't use).
>
> Yes, the build/test environment is specified by the .travis.yml file
> in
> the selinux-testsuite repo. Currently uses a bionic Ubuntu distro
> as
> the base (sadly Fedora isn't an option). In other situations where
> we
> have had dependency problems, we have explicitly had it download the
> desired upstream sources and install them, e.g. for perl tidy,
> libbpf,
> keytuils, etc. Should be possible to do the same with the kernel
> headers instead of just using the distro-provided ones. Or your
> test
> code could have #ifndef FAN_MARK_FILESYSTEM...#define
> FAN_MARK_FILESYSTEM...#endif; we have some instances of that already
> for
> SO_PEERSEC, SCM_SECURITY, MAP_HUGETLB, AF_KCM, etc. That's probably
> the
> easiest solution.
Thanks. I'll add the ifndef etc.
>
>
