[apparmor] aa-mergeprof testing results/bugs

[Christian Boltz]

Hello,

Am Mittwoch, 30. Juli 2014 schrieb Kshitij Gupta:
> On Jul 29, 2014 4:45 PM, "Christian Boltz" <apparmor at cboltz.de> wrote:
> > While testing aa-mergeprof, I found some issues:
> > 
> > - (F)inish does nothing
> 
> This was fixed for other tools(aa-genprof and aa-logprof) IIRC to
> provide expected behavior right? Maybe some tweaking is needed to
> propogate the same here.

Indeed. That should be a simple copy&paste fix ;-)

> > - args.dir is ununsed - see my other mail with a 90% patch
> > 
> > - even when (I)gnore'ing all changes, mergeprof asks to save the
> > profile> 
> >   (and behaves like cleanprof)
> 
> Hmm.. Probably a simple check if the profile was modified or not
> should suffice for unmodified profiles and Ignore use cases.

Right. (aa-logprof already does that, BTW)

> > - aa-mergeprof sometimes crashes when saving the profile, but I
> > don't
> > 
> >   have a clear reproducer yet.
> 
> Do the crash happen when trying to save unmodified profiles only or
> also when the profile is modified?

IIRC it only happened when saving modified profiles, but I'm not 100% 
sure. The worst part is "sometimes" which means I can't give you a 
reproducer (yet) :-(

We should probably check if changed[] is set in the correct way for 
every modification - I wouldn't be surprised if merging a specific rule 
type (abstractions? *x rules?) "forgets" to set changed[].


It would also be a good idea to merge ask_the_questions() in 
aa-mergeprof with ask_the_questions() in aa.py. Most things should be 
the same, the only exceptions are #include and that *x conflicts are 
possible.


Regards,

Christian Boltz
-- 
Meinem Gefühl nach hat beim systemd SuSE die Bananenpolitik
perfektioniert - nicht mal die Bananen reifen beim Kunden sondern die
Samen für die Pflanzen wurden verteilt [Manfred Kreisl in opensuse-de]