Hi Chris, I was looking at that test too - especially WRT to this comment at the beginning:
/* * This tests keep-alive behavior using chunkedinputstreams * It checks that keep-alive connections are used and also * that requests are not being repeated (due to errors) * * It also checks that the keepalive connections are closed eventually * because the test will not terminate if the connections * are not closed by the keep-alive timer. */ I think that with your proposed change then the test will no longer check that the `keepalive connections are closed eventually`. Maybe that's OK - in which case that comment should be corrected. best regards, -- daniel On 25/09/2018 10:01, Chris Hegarty wrote:
This is a test-only change to allow the test to terminate its worker threads more timely, rather than leaving it to jtreg ( which has caused some issues, see 8208690 and 7902259 ). The test now closes the worker's socket so that the worker thread, that blocked may be blocked in a socket read, can exit more timely. http://cr.openjdk.java.net/~chegar/8211092/ -Chris. [1] https://bugs.openjdk.java.net/browse/JDK-8208690 [2] https://bugs.openjdk.java.net/browse/CODETOOLS-7902259
