Github user nickwallen commented on the issue:
https://github.com/apache/metron/pull/1015
> @mmiklavc: Can you call out the lines relevant to the fix for reviewers?
I provided some more color in the description this time. Also, copied
below.
> The core problem is that the test makes assumptions about the cache state
which are not always true. Specifically, even though the cache max size is 2,
the cache can exceed that size for short periods of time until cache
maintenance is performed.
> Since the cache can be larger than expected for brief periods of time,
the expectations of what is a hit or miss were incorrect. This is why the test
would fail intermittently.
> To address this, I rewrote the test so that the cache stats are validated
after each expression is run and I also perform cache maintenance between each
execution and validation of cache state.
> Without the additional logging, the ability to grab more detailed cache
stats, and breaking down that test case, I wouldn't have been able to uncover
the bug.
---
