A bfc.C file was run over a DAQ file with 10k events, taken late 2011, which includes the FGT.  Additionally, a different bfc.C was run on 300 events from a Pythia W .fz file for the purpose of testing the FGT slow simulator.  Four difference sets of code was used, and the resident and virtual memory was recorded every 30 seconds.  Plots are shown of the resident and virtual memory size as a function of the number of events processed.  The results show no obvious leaks arising from the libraries in offline/StDevel/StRoot.  The results also verify that both data and Monte Carlo can be run with a BFC and that the FGT data is recorded in the resulting MuDst file.

The four test circumstances are

1. Starver dev,, bfc does not call any FGT makers
2. Starver eval, bfc does not call any FGT makers
3. Starver eval + offline/StDevel/StRoot libraries, bfc does not call any FGT makers
4. Starver eval + offline/StDevel/StRoot libraries, bfc calls FGT makers

The files used were

st_physics_12348038_raw_1010001.MuDst.root, taken Dec. 14 or 15, 2011
/star/data05/scratch/balewski/fgtEveGen/Wprod_b.fzd

The BFCs and .xml files (and all other details) can be found in the directories

/star/u/sgliske/FGT/2012-Jan/leak-test-MC
/star/u/sgliske/FGT/2012-Jan/leak-test-DAQ

DAQ file -- Resident Memory   

DAQ file -- Virtual Memory   

MC file -- Resident Memory

MC file -- Virtual Memory

Commentary

Note: the first few events are cut off due to the vertical scale.   Note, the shape is generally the same for all four tests.  Specifically, other than adding a constant offset, no change in the slope of the memory increase (indicative of a memory leak) is seen in any of the plots.