This is the system test report for v6r2p1.
BackGndAvg
There is something seriously wrong. The initial distributions (direction,
energy etc) of the particles looks very wrong. There are no iteractions
in the detector and no triggers. This is the only composite source in the
system tests and is also the only test that uses a precompiled source
class (Chime).
Investigations on BackGndAvg source (RXD)
I created a jobOptions file that just runs FluxAlg. I find that the energy consistently comes back as zero from Spectrum::energy(). This started in GR v6r2; worked fine in v6r1p2. I tried this both on linux and windows; the former writing out the MC root file and verifying zero energies; the latter in the debugger.
Following along in the debugger, it appears that the flat() randoms function is not passing along its return value properly.
Call stack is:
FluxSource::computeLaunch, line 667
FluxSource::event, line 635
CompositeSource::event, line 56
etc
m_energy is always zero.
I tried the default TripleRand, and HepJamesRandom generators, via GlastRandomSvc.RandomEngine. In both cases, the generated value was non-zero up to the return from the function, but was zero outside it. I'm at a loss as to where the value went. I looked around on the floor beside my laptop, but nada.
Further investigations on BackGndAvg source (JME)
I recompiled FluxSvc with GlastRelease v6r2p1, but using the previous version (v8r13) of flux. The problem went away. It seems likely that some of the recent clean up of the flux package may have cleaned a little too much...
Fix to flux (THB)
The transition from flux v8r13 to v8r14 broke CHIMEspectrum, so that it was returning NaN for energy. This is
now fixed in v8r14p1. I hope that, although the connection to the strange behavior above is not clear, this
will fix it.