Issues:
How to proceed when a module dosen't behave well (e.g. LV trip)
note : Requires modification to DSP code to get histogramming to continue when errors are encountered!Possible solutions for downstream analysis:
- Modifies IS TestData to include list of bad modules, to be added to by API
- is this sufficient for all scenarios?
- Does this need to be done on a per-test basis? Current status probably insufficient.
- API produces dummy RawScanResult with some sort of status info
- a bit nasty? (assuming something like solution (1) is used above!)
- Add bool memebers expectRawResult, expectFitResult, expectTestResult (and archiveMe?) to TestData in IS (and to TestRequest?).
- What to do if dont expect result, but get one?
- Do your stuff anyway - what have you to lose!
- What to do if expect result
- Watch IS to see if module is marked as bad
- If it is bad, dont do dependant actions (e.g. remove analysis of test if final Raw is not produced)
- Otherwise continue as usual.
- Behavour of Archiving*
- Archive control objects with list of bad.
- Various implications for DB structure need to be investigated.
Matt to add status information
Ownership/deletion of data
Problems with current solution
- General philosophy is that GUI should be farily distinct from the nasty low-level stuff.
- If two or more GUIs are running (as was the case in Oxford yesterday), could cause synch problems.
- What does GUI do if services are busy() Analysis:
- Data is really shared between many processes.
- Needs some sort of idea of whether to expect data of various types
- Some data will be e.g. file cache of retrieved data requested and used by GUI.
Suggestions
- To me it seems that clean-up on end of run is the best option.
- CalibrationController refuses to do EOR until Analysis and Archiving are done
- CalibrationController clears scratch on EOR.
- problem with circular dependancies in interfaces.
Postponed!
Viewing of ArchiveData
Largely Dave and Alan to discuss. Archiving to move to CondDB? in Release 4 - work required here.- Is it worth having an interrim solution (I'd say yes!)
- Is current ArchivingService interface sufficient?
- Awkwark bug in IS means work-around is required for retrieval of control objects. Fixed in online 00-21-02. Alan can do work-around until that is in use.
Alan and Dave to discuss
Online version
- Move to 00-21-02 is a definate possibility for release 4
Raw Scans
- Think about implications for structure/storage
- Analysis?
- Implications for CC?
- Implications for GUI
- Part of a larger issue to do with setup