Robert Muir

why do we have to make all these internal classes public to do this? I don't think this is a good tradeoff.

Yeah, those are actually API bugs? We have public methods that have non-public classes in their signature. Looks like this will be more complex to fix up. In this example...

`./gradlew check` still fails. looks like the new test catches interruptedexception and doesn't use the exception. in our codebase compiler/linter will fail on this. please annotate the variable with `@SuppressWarnings("unused")`

`./gradlew check` still fails.

Thanks for getting the tests passing. I'm honestly torn on this issue, here are my thoughts: 1. If you have 100% cpu in `java.lang.ThreadLocal.remove()`, you are creating and destroying **far...

See this is my problem, applications using threadpools of absurd sizes (unbounded, 10000, etc), and then blaming lucene for their GC woes. at this point the DSM-IV is needed, because...

The analyzers need to not be slow at query-time too. A threadlocal is a reasonable datastructure to use, you see them in every programming language: https://en.wikipedia.org/wiki/Thread-local_storage I want to see...

I do agree with removing any threadlocals on SegmentCoreReaders, if this is somehow possible to do, without annoying people who bring back 10,000 search results :) But this is a...

> Elasticsearch uses one unbounded threadpool I do find it hilarious how casually you state this. Only in java, man :)

+1 let's make a separate change to look at re-rolling the loop back. the JVM bug in question doesn't affect this version of lucene (or probably the last one either?),...