I've finally found out what my PermGen problem is being caused by. When developing the oasis tests runtime bridge some months back I put OASIS_TESTENV_RUNTIME_BRIDGE_CLASS=client.TuscanyRuntimeBridge into my environment and then forgot about it. So, that code having moved on in my absence, it turns out that I have been using the OSGi variant of the bridge recently, which exhibits this PermGen leak behaviour. I haven't resolved where the leak in that code, but I can now get nearly all the tests running.
Kelvin. 2009/11/9 kelvin goodson <[email protected]>: > thanks Simon - I'll have a go at that > > 2009/11/9 Simon Laws <[email protected]>: >> No idea why you are seeing it and I'm not. Another thing we could do >> is actually debug the perm gen failure and see what's causing it in >> your environment. >> >> Run up jconsole while the tests are running and watch the graphs rise. >> >> Assuming you just see perm gen increasing until it fails you can get a >> heap dump with >> >> jmap -dump:file=app.bin ProcessNumber >> >> And you can look inside it with >> >> jhat -J-Xmx512m app.bin >> >> Point you browser at http://localhost:7000/ >> >> There's a lot of info in there so if you do go down this route let me >> know and I can point out what to look for. >> >> Simon >> >
