Opened 17 years ago
#1288 new defect
relphot needs to be robust against NaN and should probably do flag-checking
| Reported by: | Sebastian Jester | Owned by: | eugene |
|---|---|---|---|
| Priority: | highest | Milestone: | |
| Component: | DVO | Version: | current |
| Severity: | critical | Keywords: | |
| Cc: | arminrest |
Description
When relphot is used to compute average magnitudes etc., i currently does not trap NaNs. That means an NaN in any individual measurement or error turns the entire average magnitude or error into an NaN, which then messes up colour computations.
Measurements with "fatal" flags set (e.g. saturation, CR, etc. - to be defined in detail) should probably not be included in the average, either. There is an int STAR_BAD used in some places of source:trunk/Ohana/src/relphot/src/relphot.c but it doesn't seem to be used in source:trunk/Ohana/src/relphot/src/liststats.c as far as I can tell.
See changeset [24979] from my Ohana branch for pseudocode with a suggested fix for the NaN part.
(BTW: which liststats.c is the "proper" one, the one in photdbc or relphot?)
