Re: [p2-dev] Very long planning times

[Thomas Hallgren <thomas@xxxxxxx>]

Pascal Rapicault wrote:
> The actual decision whether or not a solution exist *is* time bounded 
> (believe it is something like one minute). I have never seen the 
> solver time out on problems generated by p2.
> The explanation time is not time bounded and this is where you are 
> experiencing delay. However looking at the stack dumps attached to bug 
> https://bugs.eclipse.org/bugs/show_bug.cgi?id=274907 it seems like if 
> there is something else going on (for example on the second dump sat4j 
> is not executing...).
Yep. It's likely that the ExplanationJob fires its notify before the 
getExplanation() method enters its synchronized block. The notification 
is thereby lost and the wait is forever. I provided a patch that fixes 
that issue.

- thomas