On Fri, Apr 25, 2014 at 03:41:26PM -0400, Paul Moore wrote:
> On Friday, April 25, 2014 05:19:40 PM Markos Chandras wrote:
> > On Fri, Apr 25, 2014 at 03:07:58PM +0100, Markos Chandras wrote:
> > > On Fri, Apr 25, 2014 at 09:27:29AM -0400, Paul Moore wrote:
>
> ...
>
> > > > However, I'm more concerned that you haven't been testing on native MIPS
> > > > systems; I thought you were. We've discussed this in the past and I'm
> > > > relying on you to run these tests, as well as the "live" tests
> > > > ('regression -T live'), on the different MIPS archs/ABIs. Without any
> > > > verification of the full set of regression tests on the MIPS archs/ABIs
> > > > I will be forced to remove MIPS code from libseccomp.
> > >
> > > I didn't have the time to run the tests on big-endian systems in the past
> > > few weeks. I did back when I submitted v1 but since then a lot of things
> > > have changed. At the moment, I am focusing on making the simulator work
> > > on x86_64 hosts so most people will not notice regressions. After that, I
> > > will move into introducing live tests for big-endian systems and run the
> > > test suite in BE and LE to see what's left to be fixed. Live tests pass
> > > fine on LE MIPS last time I tried. It's only big endian that needs more
> > > testing after the recent changes, but like I said, I thought fixing the
> > > simulator results was more important than providing live tests.
>
> The short answer is that everything is important and everything must work on
> all supported architectures.
>
> The long(er) answer is that I personally would prioritize your work on the
> different MIPS host systems as that is work that likely only you can do at
> the
> moment. There are plenty of us with x86_64 systems, but not so many with
> MIPS
> systems. In hindsight, what I probably should have done is to simply
> hold-off
> and wait until the entire MIPS patchset was finished and workable (native
> tests, non-native tests, etc.); if anything this is my fault for merging the
> MIPS support to soon. I'll be smarter about things like this in the future.
It's probably a misunderstanding, or to state that better, my false assumption
that it was irrelevant where the simulator test were executed. I thought it made
no difference whether they're executed on a x86_64 or MIPS host since the
functionality of the test was tested by a simulator which I though it was
platform agnostic. I thought that only live tests were meant to be executed
on the target that's why I was not testing the simulator as much as it was
necessary. Now everything is clear and I apologize for that.
> > - Most simulator tests pass except those adding little endian architectures
> > to the filter (tests 16, 17, 19 and 23). Any prefered way to fix 17 and 19
> > tests properly or should I just duplicate them for big-endian?
>
> I can go ahead and fix those up, look for a patch and please give it a try on
> your MIPS system.
Great. I will let you know if I see problems after your patch
--
markos
------------------------------------------------------------------------------
Start Your Social Network Today - Download eXo Platform
Build your Enterprise Intranet with eXo Platform Software
Java Based Open Source Intranet - Social, Extensible, Cloud Ready
Get Started Now And Turn Your Intranet Into A Collaboration Platform
http://p.sf.net/sfu/ExoPlatform
_______________________________________________
libseccomp-discuss mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/libseccomp-discuss
