Re: [virgo-dev] Duration for resolving uses-conflicts

[Stefan Zugal <stefan.zugal@xxxxxxx>]

Hi Dani,

thanks for your reply! I tried to distill the problem and opened a Bug here:
https://bugs.eclipse.org/bugs/show_bug.cgi?id=512166

As far as I know, when resolving bundles, the OSGI framework needs to 
resolve uses-conflicts [1]. As far as I could find out, this takes place 
in org.eclipse.osgi.internal.module.ResolverImpl.checkUsesConstraints 
and produces the following debug output:

[...]
[2017-02-14 11:36:18.895] Combination is not better than current best: 
172>=172
[2017-02-14 11:36:18.895] [1,1,1,1,1,0,2,0,0]
[2017-02-14 11:36:18.939] Combination is not better than current best: 
176>=172
[2017-02-14 11:36:18.939] [0,2,1,1,1,0,2,0,0]
[2017-02-14 11:36:18.986] Combination is not better than current best: 
172>=172
[2017-02-14 11:36:18.986] Uses constraint check has timedout. Using the 
best solution found so far.
[2017-02-14 11:36:18.986] Best combination found: [0,0,0,0,0,0,0,0,0]

It seems that the resolver still has 172 conflicts, but could not find 
any better combination - so it gives up and just tries to resolve with 
the configuration given above. The same output should be reproducible 
with the setup I have attached to the Bug.

Cheers,
Stefan

[1] http://spring.io/blog/2008/11/22/diagnosing-osgi-uses-conflicts/


On 13.02.2017 21:34, dmarthaler@xxxxxxx wrote:
> Hi Stefan
>
> Could you please open a Bug on Bugzilla against 3.7.0.RC01, including 
> a simple set-up how we can reproduce this behavior? In the best case 
> you can attach an example which we can use to deploy and test.
>
> Please also elaborate a bit more about your statement: "... after a 
> certain duration the resolver gives up and just returns the 
> configuration with the least conflicts". In my experience the bundles 
> are resolved or fail to resolve, I don't know anything about a good or 
> a less good conflict resolution mechanism in Virgo. So my guess is I 
> do not fully understand what you mean by that?
>
> Regards,
> Dani
>
>
> On 02/08/2017 10:04 AM, Stefan Zugal wrote:
> > Hi all,
> >
> > I am also looking forward to Virgo 3.7.0 - thanks for all the effort! 
> > We are currently testing our software on the newest build and facing 
> > troubles with resolving bundles:
> >
> > When starting Virgo with our bundles added, everything resolves fine 
> > and Virgo starts nicely. As soon as we add the Snaps plan (which we 
> > need) to the "initialArtifacts" in 
> > org.eclipse.virgo.kernel.userregion.properties, the startup hangs for 
> > 30-90 seconds and proceeds afterwards. After some digging we found 
> > out that the time is spent in resolving uses-conflicts - after a 
> > certain duration the resolver gives up and just returns the 
> > configuration with the least conflicts. These remaining conflicts 
> > seem fine enough for resolving the bundles. When the Snaps plan is 
> > not included in the initialArtifacts, also uses-conflicts can be 
> > found, but those are quickly resolved.
> >
> > We found that the resolving of uses-conflicts can be turned off using 
> > option "osgi.resolver.usesMode=ignore" (however, that causes Virgo to 
> > fail at all). Does anybody have experience in debugging these 
> > uses-conflicts? As far as I understand, this is more an OSGI issue 
> > than a Virgo issue, but I hoped that perhaps someone of you might 
> > have an idea.
> >
> > Cheers,
> > Stefan
> >
> > _______________________________________________
> > virgo-dev mailing list
> > virgo-dev@xxxxxxxxxxx
> > To change your delivery options, retrieve your password, or 
> > unsubscribe from this list, visit
> > https://dev.eclipse.org/mailman/listinfo/virgo-dev
>
> _______________________________________________
> virgo-dev mailing list
> virgo-dev@xxxxxxxxxxx
> To change your delivery options, retrieve your password, or 
> unsubscribe from this list, visit
> https://dev.eclipse.org/mailman/listinfo/virgo-dev