Alright, branch u/hfc-3/
DM-4266 in ip_isr and tickets/ DM-4266 in obs_lsstSim have the latest corrections. No other obs_* packages call readIsrData.
The current solution is that the raw exposure is passed to readIsrData, and there is no default. Hopefully the comment of the argument is clear that it won't encourage misuse.
For now FringeTask.readFringes still returns a Struct including fringes, I am still thinking to refactor so that it returns fringes only, probably in a future ticket. It was discussed a little bit in HipChat (Nov 12, Science Pipelines room) and my notes from there:
- It is very important to be entirely deterministic
- The seed should be different for every exposure (fringe exposure?)
- Would be even better if the fringe regions and fringe measurements are persisted