Bug: moving fake plate loses light settings

[mtbcarroll]

If I import a fake plate then move it, its LightSettings instances get deleted. In looking into this, it'd be good to note if there's something odd about the fake plates' model graph when compared with real plates, and to consider when it is appropriate to delete light settings to make sure the bug isn't fixed a little too earnestly.

(There may also be a way to reproduce this without needing a plate.)

[mtbcarroll]

Anyone else should feel free to steal this ticket if they want to get a first taste of debugging graph policies, I'll be happy to help.

[mtbcarroll]

None of my usual test data seems to have LightSettings, at least that I've found so far, and I don't think any of our integration tests chgrp any either. Do we have some good test data that is known to have LightSettings? I don't see the term used in any of our readers but I'm obviously missing something if fake screens have them.

[mtbcarroll]

I found some examples of data with light settings in test_images_good: in flex/, l2d/venu/, leica-lif-martin/. If I import and move these (only the flex one is a plate) then the light settings move too. So, I looked in curated/flex/ and plates I can import from there also move fine. So, the fake reader may be somehow deficient in its plate construction.

[mtbcarroll]

At first glance my initial surprises with the fake plate are:

• its images don't reference its instrument
• one of its light source settings instances isn't referenced by a logical channel.

Is the latter "wrong" or should it be something we can cope with?

[mtbcarroll]

Judging by the naming, perhaps the DetectorSettings, LightSettings, ObjectiveSettings should always be kept with the Detector, LightSource, Objective respectively, regardless of other linkages to them. If nobody vetoes then I'll look at adjusting the graph policy rules accordingly.

[mtbcarroll]

Fixed by ​https://github.com/openmicroscopy/openmicroscopy/pull/4564.