(Small) Differences in track fit between TDR and master
After fixing the sequence in TDR for Run1 and Run2 the current tests fail with small differences in some counters related to tracking variables, e.g. one track more or less, see here https://lhcb-nightlies.cern.ch/logs/tests/nightly/lhcb-tdr-test/218/x86_64-centos7-gcc7-opt/Brunel/#Brunel.2012-bigevents
Looking closer it seems that the result of the track fit is slightly different for many tracks.
- GhostProb : 0.00379073
+ GhostProb : 0.00379072
- covariance : [ 1.58523e-05-4.84139e-07-1.24643e-07 1.23243e-08 3.31273e-12
+ covariance : [ 1.58523e-05-4.84139e-07-1.24643e-07 1.23243e-08 3.31274e-12
- 3.31273e-12 1.69669e-13-1.69898e-14-2.60849e-15 5.33331e-15 ]
+ 3.31274e-12 1.69669e-13-1.69898e-14-2.60849e-15 5.33331e-15 ]
459c460
- 3.34376e-08 1.61574e-07 9.34264e-11 1.70389e-09-1.08502e-14
+ 3.34376e-08 1.61574e-07 9.34264e-11 1.70388e-09-1.08502e-14
469,470c470,471
- 4.32608e-08-5.86457e-09 1.11769e-09-2.37579e-12-1.19017e-13
- 1.66975e-08 3.94698e-07-2.37579e-12 3.18365e-09 5.70678e-14
+ 4.32608e-08-5.86457e-09 1.11769e-09-2.37578e-12-1.19017e-13
+ 1.66975e-08 3.94698e-07-2.37578e-12 3.18365e-09 5.70678e-14
490,491c491,492
- 0.000750795 0.0396896 -3.4508e-07 6.08625e-06 8.0097e-10
- -4.82628e-07 -3.4508e-07 1.43454e-09-6.02085e-11-2.45557e-12
+ 0.000750795 0.0396896-3.45079e-07 6.08625e-06 8.0097e-10
+ -4.82628e-07-3.45079e-07 1.43454e-09-6.02085e-11-2.45557e-12
502c503
- 7.27281e-07-8.39377e-07 2.81887e-09-9.99651e-11 2.46804e-14
+ 7.27281e-07-8.39377e-07 2.81887e-09-9.99651e-11 2.46803e-14
The differences are usually on the fifth digit.
What should we do? Update the references or try to find the difference (if possible).