Community
Participate
Working Groups
WTP 1.0M4 Scenario: Run through the bottom-up Web service creation/client scenario, all the widgets/commands/etc are leaked. As a result, all the things that they pull in, for example, WSDL model, J2EE models and etc are leaked as well.
Here's an update: I've verified that there are objects persisting in the JVM after the wizards are done that should not be around. As the cause (or anchor) could be any number of objects, I will open individual bugs to target certain objects in packages which may be a potential culprit. Performance/Memory leak-related bugs: https://bugs.eclipse.org/bugs/show_bug.cgi?id=98709
Using the profiler, I saw that most of the jst.ws* objects being held in memory were rooted in(for us)the Dynamic Popup Action launch point (in the case of pop- up menus). And in the bigger picture, the above class is being held by the Display and MenuItem classes. I'll investigate how to remove this link and free it up for GC. Any ideas would be appreciated.
With the fix made by Peter for 98709, I was able to measure a big decrease in memory leakages from wst.command and jst.ws. Some ad-hoc testing with a profiler showed memory reduction from 2MB to 500KB. Once the next integration driver with the fix is declared, I plan to do more systematic tests. In the meantime, we should be able to reduce the severity of this bug to normal based on the initial results. More fixes are planned to reduce the memory footprint in M6, so we can leave this bug to track this. If there are any objections or comments, please let me know.
Deferring the remainder of this bug to WTP 1.0.
Deferring bugs of normal or lower severity out of WTP 1.0 per [1]. Please let me know if you think this bug should be reconsidered with a higher severity for WTP 1.0. [1] http://dev.eclipse.org/mhonarc/lists/wtp-dev/msg02831.html
We fixed the leak in https://bugs.eclipse.org/bugs/show_bug.cgi?id=98709 and left this bug open because we suspected there may be a small leak remaining. However, I recently profiled these scenarios using Yourkit Java Profiler and found there is no leak. I compared snapshots before running each scenario and again after running the scenarios at least 5 more times. No memory growth in the jst.ws* or wst.ws* was seen. I'll be resolving and closing this bug as invalid. Let me know if there are any questions.
Profiled WTP v1.5RC1
Done