| Re: [tracecompass-dev] Call stack view not sorted correctly by depth |
Okay, here's what I have learned:When I first open the trace in the call stack view, it is sorted correctly. If I leave the call stack view open, but choose "Delete Supplementary Files..." from the trace context menu and delete all of the files, then it will reindex and run the analysis again - at this point the sorting is out of whack. If I close the call stack view and reopen it, the sorting is correct. So, some state must be saved in the view itself that is not being wiped when deleting the supplementary files. Does that seem plausible? I can consistently reproduce the issue.Now, maybe most users do not constantly delete and reindex, but it is common for me as a developer...RockyOn Tue, Oct 3, 2017 at 11:34 AM, Rocky Dunlap <rocky.dunlap@xxxxxxxx> wrote:Thanks for all the suggestions! A simple restart looks like it has resolved the problem. I'm going to keep testing a bit more just to be confident, but I hope it was perhaps due to a corrupt state system from using the debugger...RockyOn Tue, Oct 3, 2017 at 11:31 AM, Jonah Graham <jonah@xxxxxxxxxxxxxxxx> wrote:Hi Rocky,I had a similar problem a while ago due to problems in my event stream. I had warning messages I didn't see at first (of the form "Function exit name in event ({0}) different from the expected one ({1}). You may have lost events in your trace", message from CallStackStateProvider.eventHandle(ITmfEvent)). Is there a warning in your log along those lines? If so, you may need to look at your data source. Jonah~~~
Jonah Graham
Kichwa Coders Ltd.
www.kichwacoders.comOn 3 October 2017 at 18:16, Patrick Tasse <patrick.tasse@xxxxxxxxx> wrote:The children of ThreadEntry should never be rearranged by a sort, as the CallStackComparator always returns 0 when comparing CallStackEntry's.Hi Rocky,As far as I can see from the source code, instances of CallStackEntry are assigned their stack level from an internal counter that is incremented as each new child is added to the parent (ThreadEntry in your case).Not sure this helps, but if you notice anything that appears incompatible with the above, let us know...PatrickOn Tue, Oct 3, 2017 at 12:06 PM, Rocky Dunlap - NOAA Affiliate <rocky.dunlap@xxxxxxxx> wrote:______________________________I have implemented a custom call stack analysis by extending CallStackStateProvider. In one trace I have a call stack depth of 33. When I view it in the call stack view, it is for some reason not sorted by depth. Notice in the screenshot that the values in the depth column itself are correct, but the stack frame UI elements are not sorted visually by this depth. There is some kind of interesting pattern that might shed light -- the even numbered depths are at the bottom (except for 1) and the odd depths are at the top.Any ideas what could lead to this or what I might be doing wrong?Thanks,Rocky_________________
tracecompass-dev mailing list
tracecompass-dev@xxxxxxxxxxx
To change your delivery options, retrieve your password, or unsubscribe from this list, visit
https://dev.eclipse.org/mailman/listinfo/tracecompass-dev
_______________________________________________
tracecompass-dev mailing list
tracecompass-dev@xxxxxxxxxxx
To change your delivery options, retrieve your password, or unsubscribe from this list, visit
https://dev.eclipse.org/mailman/listinfo/tracecompass-dev
_______________________________________________
tracecompass-dev mailing list
tracecompass-dev@xxxxxxxxxxx
To change your delivery options, retrieve your password, or unsubscribe from this list, visit
https://dev.eclipse.org/mailman/listinfo/tracecompass-dev
_______________________________________________
tracecompass-dev mailing list
tracecompass-dev@xxxxxxxxxxx
To change your delivery options, retrieve your password, or unsubscribe from this list, visit
https://dev.eclipse.org/mailman/listinfo/tracecompass-dev
_______________________________________________
tracecompass-dev mailing list
tracecompass-dev@xxxxxxxxxxx
To change your delivery options, retrieve your password, or unsubscribe from this list, visit
https://dev.eclipse.org/mailman/listinfo/tracecompass- dev
