[Clas_offline] problem with gpp

[Hyon-Suk Jo]

Hi everyone,

I've been running simulations, with all executables built a few months ago 
with source codes from SVN, and obtained an extremely low number of 
reconstructed events after my PID cuts (nothing comparable with what I 
used to get with my old executables). After a bunch of tests, I clearly 
identified gpp as the cause of the problem. I reprocessed my simulations 
with two different old gpp executables built in 2008 and 2009 on Fedora 8 
and obtained reasonable results. When comparing the ntuples produced from 
the simulations processed with my new gpp executable with the ones 
produced from the same simulations but processed with an old gpp 
executable, I found that while the distributions of number of 
reconstructed particules per event are similar, the id of 94% of trigger 
electrons is 0 instead of 11 when using the new gpp (it's 20% when using 
the old gpp). Then I built a new gpp executable on the current operating 
system using a 2008 revision of the source code and it gave results 
similar to the ones obtained with the old executables. So the problem 
seems to come from changes made to the code since 2008. I did a rather 
quick comparison of the current gpp code with the 2008 version and found 
no obvious problem. I could keep looking into it but it would be more 
efficient if experts of the gpp code and particularly the people who made 
modifications to the code during the past couple of years could 
investigate the problem.

PS: by the way, I tried the latest version of the source code from CVS 
(just to check) and while it is slightly different from the current SVN 
version, it gaves the same problem.

Cheers,
Hyon-Suk