Study if PVs with failed fit should be handled
When !4468 (merged) is in, the fit iteration is stopped when matrix inversion fails. According to Wouter, that's extremely rare :) but should be checked still and might be handled by removing the PVs.
The following discussion from !4468 (merged) should be addressed:
-
@wouter started a discussion: (+5 comments)
One small problem with this solution is that the fit will not abort but just keep running until the maximum number of iterations is reached. (The update now does 'nothing': so it will fail again in the next iteration.) One alternative solution is to extend the set of return values, for instance [not-converged=0, converged=1, inversionfailure=2]. The fit loop can then abort in case of inversion failure.
(Inversion failures will be extremely rare.)