STAR   Computing Tutorials main page
How to Test Changes to Reconstruction Code
  Maintained by Kathy Turner

From email sent to starsoft-l on 27Dec1999:

In the PWG review session on Dec. 14, 1999,
  --> /STAR/html/all_l/html/pwgc_summary_dec99.html
it was noted (look under QA) that people are now encouraged to
check the results of changes to the reconstruction code (here I
define as anything used to run "bfc") BEFORE submitting the changes
to CVS.  This is to speed code development and improve library quality.

The tests are to be done by checking the QA-histograms with your code
changes against the previous library versions.
Even though these tests are done every day (DSTs generated and then
the histograms made and put on the auto-QA page), there can still be
problems when there is more than 1 check-in per day.  Also, there is a
time-lag between checkin and when the DST QA-results get reported.  

The easiest way to test is to run the standard bfc and then make the histograms.

  1. stardev
  2. run root4star and execute bfc with appropriate geometry:

    You could also run the /hadronic_cocktail/lowdensity or /highdensity

    The above will execute 40 events with the standard year 1b or 2a geometry and TFS with the hadronic cocktail standard input file.
    Output files will be generated in your area, including a gstar.hist.root file.

  3. Now to send the QA histograms to a file, run root4star and
      .x bfcread_hist_to_ps("gstar.hist.root")

    You will get the "QA" histograms in a file --> QA_hist.ps

  4. Now compare your histograms with the ones made last night and kept in the auto-QA WWW database. Make sure you are comparing the histograms with same geometry and input file (lowdensity,standard,etc)

    Go to /STAR/devcgi/qa/QA_main.pm and then

  5. If there's any question that your code changes could affect the results on Solaris & Linux differently, then you should do the above steps and a Linux & a Solaris machine.