[
Date Prev][
Date Next][
Thread Prev][
Thread Next][
Date Index][
Thread Index]
[
List Home]
|
Re: [tracecompass-dev] Trace Compass Mars.1 release preparation
|
On 2015-08-14 09:46 AM, Oberhuber, Martin wrote:
Thanks for the detailed information !
I find dealing with "lost events" really important to ensure user's trust in the tool, and I really like the visualization metaphor that the tracecompass team has chosen.
Maybe one additional enhancement to consider would be teaching users how to deal with lost events.
For example, when uploading / opening a trace, a dialog could pop up
"
This trace indicates that some events have been lost.
Therefore, analysis may be incorrect in areas indicated in red.
To avoid lost events, either reduce event frequency (by collecting less events),
or increase buffer size on the target, or save to a faster file system.
[x] Don't show again [OK]
"
That's a very good point. This is something we have discussed with some
users previously too. As soon as you have lost events in the trace, all
bets are off, and it's possible that all the analysis all wrong and
meaningless.
We agreed that making it obvious that there are lost events is very
important. Where we were not too sure is how to present it to the user.
A popup is an intuitive idea, however it may becoming annoying if the
same trace opens the same popup every single time.
The "[x] Don't show again" would help, but the danger is that if you
check it, then you don't get the warning for future traces that might
also contain lost events.
Another idea was to have a message, maybe in red, in the status bar at
the bottom, that would mention that the trace has lost events. It would
hopefully be visible but not interfere with the rest of the application.
Finally, there is (planned or ongoing?) work on adding a red "overlay"
on top of all time graph views to show areas where events have been
lost. A bit like the Histogram currently does, but just to show the
intervals of time, not the actual numbers. This would be another means
of showing users that the trace has lost events, and that all results
shown during of after these periods might be wrong.
Cheers,
Alexandre
