For anyone interested. The commit involves addressing two needs. 1) Enforcing MVEL reflection mode throughout - MVEL ASM optimiser can NEVER be used for anything (it doesn’t add much value anyway). 2) Avoiding giving the template system the root class loader(which we use for dynamic stuff), as those templates are statically cached. https://github.com/sotty/drools/commit/1b75d4785861e72338cc5ea1280610a7937be017
Mark On 8 Jan 2014, at 00:53, Mark Proctor <mproc...@codehaus.org> wrote: > The root of the perm gen leak has now be solved, thanks to some heroic work > by community developer Davide Sottara. > > This is just in time for the 5.6 release going out this week. You will not > need to use the kbase.dispose() work around. > > Mark > On 6 Jan 2014, at 20:50, Mark Proctor <mproc...@codehaus.org> wrote: > >> The 5.6 approach should be considered a work around, not a fix. Somewhere a >> ref is being held, that shouldn’t be - we just haven’t found it yet. >> >> Mark >> >> On 6 Jan 2014, at 19:17, Davide Sottara <dso...@gmail.com> wrote: >> >>> I have run pmander's tests, and reported in a previous email. >>> The leak in 5.x is due to the nature of the composite classloader, and >>> dispose() >>> forces the memory to be released. I don't remember if that fix was in >>> 5.6.CR1 >>> (probably not), so you may have to try 5.6.0-SNAPSHOT or wait for the >>> end of the week. >>> 6.x does not suffer from the same problem, but has a different issue - >>> no way >>> to unload kie modules once they are cached - which Mario has been >>> working on. >>> In either case, pmander's test works fine in both 5.6.x and 6.x after >>> these fixes >>> Davide >>> >>> On 01/06/2014 07:56 PM, Mark Proctor wrote: >>>> I’ve been told 5.6 will be released this week. This is because the >>>> JCR2GIT migration tool, in 5.6 is needed for our 6.0 product. >>>> >>>> Davide's current kbase.dispose() method, on the concrete class, should >>>> just be considered a work around for the problem. Unless this is fixed by >>>> someone in the community, I don’t think any core developers will have time >>>> to address this. At least not for 5.x; we’ll continue to look into it for >>>> 6.x. >>>> >>>> Mark >>>> >>>> >>>> On 6 Jan 2014, at 16:14, brachi <brach...@sapiens.com> wrote: >>>> >>>>> see previous messages, pmander attached a unit test >>>>> >>>>> >>>>> >>>>> -- >>>>> View this message in context: >>>>> http://drools.46999.n3.nabble.com/permgen-leak-tp4027038p4027539.html >>>>> Sent from the Drools: User forum mailing list archive at Nabble.com. >>>>> _______________________________________________ >>>>> rules-users mailing list >>>>> rules-users@lists.jboss.org >>>>> https://lists.jboss.org/mailman/listinfo/rules-users >>>> >>>> _______________________________________________ >>>> rules-users mailing list >>>> rules-users@lists.jboss.org >>>> https://lists.jboss.org/mailman/listinfo/rules-users >>>> >>> >>> _______________________________________________ >>> rules-users mailing list >>> rules-users@lists.jboss.org >>> https://lists.jboss.org/mailman/listinfo/rules-users >> > _______________________________________________ rules-users mailing list rules-users@lists.jboss.org https://lists.jboss.org/mailman/listinfo/rules-users
