Community
Participate
Working Groups
See defect 196013, step 8, Open Execution Flow. Under 'Thread name'->main[xxxx]->main(java.lang.String[]), you should see an entry "unknownXXXXX()", where XXXXX is a set of numbers (in this case 69256). The unknown entry comes from MethodEntry and MethodExit of a method that does not have a MethodDef. The method is called during the pause period and so it's not sent to WB.
Please see my comment 10 to bug 196013. It may turn out to be a duplicate of bug 190687 or bug 193304.
Resolving as dup of 196013 as partial fix for 196013 was not delivered in 4.4.0.1 *** This bug has been marked as a duplicate of bug 196013 ***
Joana, we have determined that this bug is not a duplicate of 196013. The root-cause is that classDef and methodDef information is not properly cached during "pause" and therefore not transmitted upon "resume". It would be better to leave this bug open so we can track it independently of bug 196013 (which was originally submitted to address an NPE in the workbench).
Okay. I am changing the target to 4.4.1. If a 4.4.0.2 release is approved, this defect will be targeted to 4.4.0.2.
Note that in the scenario mentioned in bug 196013, 'unknown' is not shown in view if I start the Java proccess with vm options -agentlib:JPIBootLoader=JPIAgent:server=controlled;CGProf:execdetails=true instead of -agentlib:JPIBootLoader=JPIAgent:server=controlled;CGProf
The problem is identified. CGProf caches thread/method entries/exits only in aggregated mode (aggregated mode is switched on by 'execdetails=true') Guru, Asaf: Should this feature be addressed before new pause/resume logic will be available? PS: Reassigning bug to Alex, cause I'm on vacation next week.
(In reply to comment #6) > The problem is identified. > CGProf caches thread/method entries/exits only in aggregated mode (aggregated > mode is switched on by 'execdetails=true') IMHO, caching of method entry/exit events is not relevant for this defect. What needs to be cached is "classDef" and "methodDef" elements for new methods/classes identified while data collection is paused. These elements are constructed in the "New Method" event handler, not in method entry/exit event handlers. This bug should be implemented as part of applying the updated design for supporting attach/detach/pause/resume use-cases.
*** Bug 199623 has been marked as a duplicate of this bug. ***
Isn't reproduced at 4.4.1-200708280100
Resolving since not reproducible
closing
