Community
Participate
Working Groups
Build Identifier: 20110526-1053 If a trace is big enough, selecting another trace/experiment while the big one is loading will result in wrong statistics for the big trace that has been interrupted. The trace does not need to be that big. If done fast, it will result in the same misinformation. The plug-in remember that the trace was already loaded before, so it doesn't reload it, but it was not completed in the first place. Reproducible: Always Steps to Reproduce: 1. Open a big trace (1) 2. While the trace you have selected is being loaded in background, select different trace (2). 3. Get back to the big trace (1) 4. The statistics will not show the real event number of the trace (1). There will be missing events, those that were not yet loaded when you clicked on the trace (2).
The eventHeap is getting flushed before the first experiment has been disposed. Therefore, there is a NullPointerException raised.
Created attachment 200730 [details] Temporary fix to handle failed requests Temporary fix to handle failed requests while we fix the experiment concurrency issue.
Created attachment 202174 [details] Dispose correctly of the loading experiment when another one is selected Patch generated from git in command line.
(In reply to comment #3) Thanks Mathieu for the patch. I'm going to include the fix. /Bernd
Created attachment 202334 [details] Fix for bug Fix for bug. It augments attachment 202174 [details].
(In reply to comment #5) Patch is based on master.
Created attachment 202350 [details] Fix for bug (stable-0.8) Fix for bug based on branch stable-0.8.
Patches commited!
Delivered with 0.8.1
