paulirwin commented on PR #1222: URL: https://github.com/apache/lucenenet/pull/1222#issuecomment-3744880697
The PR does not need to be reverted. If this is an issue in .NET 10, it would surface to users anyways, so better to get it merged so more people can test it. I am not convinced this is a systemic concurrency problem. I ran the test locally tens of thousands of times on macOS and Ubuntu under .NET 10 and could not reproduce, even tweaking the parameters like the number of threads and wait time. If it were systemic, it would be reproducible at least once every few thousand tries. My current working theory is that .NET 10 is _so fast_ that we're now dealing with microsecond race conditions that we weren't before, and as noted in #1233, the test does not test real-world conditions. It also might be true that my machine is too fast for this to surface, and it only occurs in the significantly slower Actions containers. Even if someone were trying to create and destroy IndexWriters without reusing them that fast under extremely heavy small-write concurrent load with a RAMDirectory in the real-world, they probably would have their system resilient enough to retry on such an exception, and the chance of it happening twice is very small. But, I am in favor of #2 above. We should keep testing for this and mark awaits fix if needed. I created PR #1234 to catch the file not found exception, but if someone can demonstrate that there is an actual real-world concurrency issue here with a different solution, I can abandon that PR. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
