Envy/Stats doesn't recognize its own Samples and Traces

[jtu...@objektfabrik.de]

Try the following in a fresh VAST image (I've seen the problem in VAST 7 and 8.5).

1. Load Envy/Stats map
2. Open Performance Workshop
3. Sample or Trace a bench method
4. Try to open the Method distribution or method execution details (in the Results menu)

I have not yet found a pattern under which exact circumstances this happens, but chances are, VAST will not open the above-mentioned windows and claim that there is no sample or trace data available.

This happens with bench methods provided with Envy/Stats as well as with self-implemented bench methods (we've tried on several machines, with VAST 7 and VAST 8.5). The same bench method can be browsed in another newly started fresh image or on another user's machine without any problems, but even tracing it again in an image where the problem appears will not work. But other bench methods may be working in the same image. As I've said, no pattern so far.

Has anybody seen this behaviour and found the reasons for it? I haven't debugged into #buildTree yet because we found the problem during a workshop where I was supposed to demo the tool ;-). The nicety of this is that my demo benchmark methods had been running on the same laptop in another image just the other day. We only then tried on other machines in the classroom and saw that the same problem can affect any bench method in several images on different machines....

Joachim

[jtuchel]

Any news on this one? Any ideas?

--
View this message in context: http://forum.world.st/Envy-Stats-doesn-t-recognize-its-own-Samples-and-Traces-tp4031659p4128001.html
Sent from the Instantiations mailing list archive at Nabble.com.

[SebastianHC]

Hi Joachim,

I can't reproduce your issue.

I just took a new Image, I loaded as you said. I traced every test and
everytime I get a proper result of distribution and execution times...

The only thing I get is an "undefined object does not understand
increment:".
This occurs while tracing EsbTestBench->EsbExecutionTreeRedraw with Runs:10
and Iterations:10

But I don't think that's what you ment.

Cheers!
Sebastian

--
View this message in context: http://forum.world.st/Envy-Stats-doesn-t-recognize-its-own-Samples-and-Traces-tp4031659p4165822.html

[jtu...@objektfabrik.de]

Hi Sebastian,

strange. I hadn't seen the problem for years (actually never before), but once I had it, I saw it quite frequently, and one of the attendees in my workshop also saw it, in different Bench methods.

Next time I see it, I will have to debug the code and see what I find out. I was just hoping for somebody else to have seen it before. Obviously, I'll have to go where noone else has been before ;-)

Joachim

[John O'Keefe]

Joachim -

 

OK, I've now seen the problem (even a couple times), but now I no longer can see it (since I wasn't keeping track of exactly what steps I was running when I saw it). If you can reproduce the problem and give me the scenario, I will look into it further.  I will also continue to try to reproduce the problem.

 

John

[jtu...@objektfabrik.de]

Hi John,

good to hear I'm not hallucinating. As I said, I've found no pattern so far. I'll do some extra samples over the next few days to see if I hit the problem again. 

Joachim