| | 7 | |
| | 8 | * various relphot issues: |
| | 9 | * very bright stars (needed for ghost analysis) were failing to get reasonable magnitudes. When very saturated, these stars get broken up into multiple fainter PS1 magnitudes. My recipe for deciding to use the synthetic vs PS1 data was failing in these cases. I've made a short term fix to fall over to synthetic mags if the synthetic mags imply a very saturated star. |
| | 10 | * errors in the presence of ubercal data were getting screwed up. I was up-weighting ubercal values by simply dividing their errors by a factor of 10, but I failed to account for this in the weighted error calculation. I re-worked that code to have a separate weighting vector so that the error is calculated correctly. |
| | 11 | * there were areas of the sky which were not seeing the results of the relphot analysis. I added some per-remote client logging to track this down and found a couple of issues. First, one file on ipp008 had a hung lock, and this was crashing not just ipp008 jobs, but other ones as well (due to client / server interactions). I cleared the lock, but I'm not certain how to catch this in the future (the locking system just times out). Second, the remote clients were failing to load the image subset table because it was not yet visible over NFS. I added a flush and fsync to force the data to disk, and this helped. |
| | 12 | * started tests of the parallel dvomerge by running dvomerge on the WISE and SuperCOSMOS database. |
