PhysicalTable uses getAvailableIntervals

[archolewa]

--PhysicalTable::getAvailableIntervals returns
availableIntervalsRef.get(). However, the other methods didn't actually
use getAvailableIntervals internally. Instead, they used
availableIntervalsRef.get() directly. As a result, overriding
getAvailableIntervals would not actually be sufficient to restrict a
PhysicalTable's view of its availability, despite the method name.

--Now, all invocations of availableIntervalsRef.get() have been
replaced with getAvailableIntervals, making it much easier to enforce
a consistent view of availability simply by overriding
getAvailableIntervals.

[cdeszaq]

👍 (though a changelog entry would be nice, since this could break someone if they were relying on internal things not calling that getter)

[archolewa]

@cdeszaq The only way I can think of that happening would be if someone is overriding and changing getAvailableIntervals not to affect Fili's behavior (since we don't actually use it anywhere but tests), but solely for use in their own code, which would be a whole new level of weird.

But I've added a changelog entry anyway, because whatever.

[michael-mclawhorn]

👍
I would actually regard this as safe not to CL, although this is a perfect candidate for establishing guidance around what is and isn't a public api.

[cdeszaq]

My 1st thought was that this was below the level of noting in the CL, but since PhysicalTable is one of the Core Domains within the system, and it seems that the thrust of the PR is to make it easier to extend, it made sense to err on the side of being overly communicative in the CL about it than surprise possible extenders.