Highlights
- Intent is to run system tests before accepting any new GlastRelease releases for use in L1
- Run under the glastsys SLAC Unix account
- Results should be reviewed by subsystem "experts" - now it should be C&A
- System test code has been around for a long time, and has had a variety of care-takers Original Documentation
Notes from Simone
I managed to run the system tests to compare GR_20-04-10 and GR_20-11-00 (at least for a large sets of simulations/reconstruction). The results can be seen on the web interface.
- The code I am using is in /nfs/farm/g/glast/u17/systests/ (SLAC unix account: glastsys)
I had to update some scripts (changing some paths, and other minor things), and everything should be in CVS.
I summarized the procedure I used to run the test in this confluence page.
- There is still one issue about some ACD quantities: for the ACD recon variables we have to use the "new" (pass8) AcdReconV2 class (instead of AcdRecon). I converted some of these variables and commented out all the others (for which the old/new correspondence is not clear to me), so now some acd histograms are empty. The related code is in file systests/src/root/ReconAcdFragment.cxx
Furthermore some sets of simulations/reconstructions are not working (AllGammaOverlay, BackGndMixDC2, BackGndMixDC2Overlay ). I suppose this is related to the fluxes definitions and path to overlay files...
- Finally, when comparing GR_20-04-10 and GR_20-11-00 I found that all the tests are failing for two histograms (CALRMSTRANS and CALRMSLONG), corresponding to the recon variables ReconEvent.CalRecon.CalCluster->getRmsTrans() and getRmsLong(). I don't know if this is something expected. I also tried to run the comparison staring from the digi file of a LPA run (instead from a simulation), getting the same result (flightData row in the systests page). Digging a bit I found that in one case (GR 20-11-00) these two variables are always 0.
I summarized the procedure I used to run the test in this confluence page.
- There is still one issue about some ACD quantities: for the ACD recon variables we have to use the "new" (pass8) AcdReconV2 class (instead of AcdRecon). I converted some of these variables and commented out all the others (for which the old/new correspondence is not clear to me), so now some acd histograms are empty. The related code is in file systests/src/root/ReconAcdFragment.cxx
Furthermore some sets of simulations/reconstructions are not working (AllGammaOverlay, BackGndMixDC2, BackGndMixDC2Overlay ). I suppose this is related to the fluxes definitions and path to overlay files...
- Finally, when comparing GR_20-04-10 and GR_20-11-00 I found that all the tests are failing for two histograms (CALRMSTRANS and CALRMSLONG), corresponding to the recon variables ReconEvent.CalRecon.CalCluster->getRmsTrans() and getRmsLong(). I don't know if this is something expected. I also tried to run the comparison staring from the digi file of a LPA run (instead from a simulation), getting the same result (flightData row in the systests page). Digging a bit I found that in one case (GR 20-11-00) these two variables are always 0.