I should have a fix shortly. Jacopo
On Feb 26, 2010, at 7:31 AM, Adrian Crum wrote: > --- On Thu, 2/25/10, Jacopo Cappellato <jacopo.cappell...@hotwaxmedia.com> > wrote: >> On Feb 26, 2010, at 12:27 AM, Adam >> Heath wrote: >> >>> Adrian Crum wrote: >>>> Jacopo Cappellato wrote: >>>>> On Feb 25, 2010, at 7:54 PM, Adam Heath >> wrote: >>>>> >>>>>> Jacopo Cappellato wrote: >>>>>>> Hi Adam, >>>>>>> >>>>>>> On Feb 25, 2010, at 5:37 PM, Adam >> Heath wrote: >>>>>>> >>>>>>>> Jacopo Cappellato wrote: >>>>>>>>> (moving to the dev list) >>>>>>>>> >>>>>>>>> Hmmm.... >>>>>>>>> >>>>>>>>> I am trying to find out more >> details using jmap and now I am no >>>>>>>>> more thinking that birt is the >> issue. >>>>>>>>> Adam, is it possible that the >> Webslinger component is causing this? >>>>>>>>> Running jmap -permstat I get >> the following stats: >>>>>>>> Not the component, no. What >> you see below is a consequence of the >>>>>>>> custom classes that get compiled >> for every single service engine >>>>>>>> definition, so that you can see >> what file/line the service is >>>>>>>> defined in. >>>>>>>> >>>>>>> sorry for the silly question: are you >> saying that the custom classes >>>>>>> are not compiled by some code in >> Webslinger? I really don't know >>>>>>> where to look for AsmUtil and >> GeneratedClassLoader. >>>>>> I never said any such thing. I'm >> just saying that we may need more >>>>>> permgen space, as that is where classes >> are loaded. >>>>> >>>>> Is there a way to temporarily disable this >> GeneratedClassLoader thing? >>>>> Could you point me to the right direction for >> fixing this issue >>>>> (without the hack of increasing the PermGen >> memory)? I could not even >>>>> find the AsmUtil class... >>>> >>>> Jacopo, it looks like the code you are wanting to >> work on is contained >>>> in the webslinger jar files. We don't have access >> to the source code. >>> >>> Yes, this is true, but the interface to that is in the >> service engine, >>> so you could modify ModelServiceReader, as I suggested >> previously. >>> >> >> Ok, thanks guys, I'll have a look at this. > > Good luck! I looked at it and gave up. ;-) > > > >
