IssueTracker

Thanks for the thorough report - you're right about the problem, and the workaround looks good - it even handles the fast cases for initialization and setList(null) nicely.

Fix and tests submitted internally, should go out with next paging release.

Actually, I'd like to know more about what failed here and why.

I'm curious if InstantTaskExecutorRule is putting paging into an unexpected state - do you have a sample project that reproduces the issue?

When hitting this outside of a test, are you specifying any custom executors, possibly ones which run main thread tasks immediately instead of posting them?

I don't have a sample project that hits this.

The problem does not happen very often outside of the test scripts, but we are not running any custom executors. The PagedListAdapter gets the data through a LivePagedListProvider from room db.

This specific issue should be fixed with the Paging Alpha 4 that just released.

Filed https://issuetracker.google.com/issues/70351983 to track testing around InstantTaskExecutorRule.

Indeed I can reproduce the issue with a simple project (see attachment) where the base class and the child class are in 2 modules.  If the annotationProcessor is commented, it works as expected.  If it's enabled, only the child class method is called.

Thank you, now I can reproduce it, looks we doesn't correctly different modules in our annotation processor.