On Thu, Oct 01, 2015 at 08:44:11PM +0200, Jakub Hrozek wrote:
On Thu, Oct 01, 2015 at 08:39:15PM +0200, Jakub Hrozek wrote:
> On Thu, Oct 01, 2015 at 08:31:50PM +0200, Lukas Slebodnik wrote:
> > On (01/10/15 20:13), Jakub Hrozek wrote:
> > >On Fri, Sep 25, 2015 at 07:49:23AM +0200, Lukas Slebodnik wrote:
> > >> There is a common agreement (at least from management side) that we
want
> > >> to improve code coverage This patch is great opportunity to add
regression
> > >> test. You can choose either integration test or unit test; you should
know
> > >> what suits better to this case.
> > >
> > >I tried to create a test (see my tests branch at fp.o) but it's really
> > >not easy, the sdap_async.c module has quite some dependencies and the
> > >generic search request is real pita to mock. And we need to apply this
> > >patch to downstream.
> > >
> > >So while I agree with you we need tests, I think this one needs to be
> > >done together with the LDAP refactoring, sorry.
> > What about integration test?
>
> Calling sdap_posix_check_send requires LDAP backend with ID mapping and
> AD schema..
Actually ID mapping should be disabled of course...
So what we could do is start a test while the server only contains entry
with a wrong POSIX data to make sure the request hits it and then add a
correct entry and request it..
Turns out that the test hits the POSIX check, but doesn't error out
with:
id_provider = ldap
ldap_schema = ad
Because the code that triggers the next connection (and eventually marks
sssd as offline) is only in the AD provider.
The WIP test can be found here:
https://fedorapeople.org/cgit/jhrozek/public_git/sssd.git/commit/?h=intg_...
So we really need Samba4 AD to test this..sorry, I'll push the patch w/o
tests.