From: Julie McEnery
Sent: Tuesday, December 07, 2004 5:34 PM
To: Dubois, Richard
Cc: Julie McEnery
Subject: systests
Hi Richard,
I am happy with the entrance to the system tests - this is the page with several boxes allowing one to select which GlastRelease or EngineeringModel version to look at. We probably need only one submit button.
The next page (which lists the list of tests for a given release):
It would be great to reformat this a little. In particular it would be better to reformat things so that each line gets narrower, so that more tests can be viewed in the same screen (this could be done by altering the "plot histos get tuples" box.)
In the datafiles section, it would be better to make this be a link to the log file and the merit tuple, and the systest name could lead people to the plot histos interface. It would be nice to find a way to show people how to download the raw root files from ths page (they are always in the same dir as the tuple).
In addition to being able to change the threshold value, it would also be good to be able to change the standard and resubmit and recalc the pass/fail values on the fly. This would be nice but I don't consider it a priority. One could also put a button here to change the test that decides on pass/fail (i.e. to use something other than KS).
The status of pending/failed jobs is not always correctly handled.
If the status in the system_tests table is pending or failed then the row should be greyed out (possibly different shade for the two cases).
If the entry is "NA" in the histogram_file field, then there should not be a link to "plot histos".
It would be nice to actually see the histos that fail when one clicks on the fail button. Would it be possible to display them (maybe with a limit on how many per page) in addition to the meta data that is already displayed? It would be great if these plots were clickable and that clicking on one would lead you back to the single plot page where one could do more fiddling.
It would be nice to pick up and use the axis labels associated with each root histo when making the plots. It would also be nice to display the values of the under/overflow bins. It would also be nice to see the values of the statisical comparison test(s). Can we add a facility to rebin the histos (into bigger bins), something akin to what you get with a TH1F::Rebin() in root?
These are the things that are mostly an extension of the things that are already there. The other things to consider is a (much more major) revamp of the plot histos page, and adding a method to flag histos as being different but deemed not a problem (so that we can move towards a method for declaring a release to have passed or failed the systests). I'll continue with my thoughts on these in the next email.
Regards,
Julie