On Mon, 13 Jan 2025 16:07:45 GMT, Daniel Fuchs <dfu...@openjdk.org> wrote:
> There are a few places in the HttpClient code base where we close the > connection when some error/exception happens. This can some time trigger a > race condition where closing the connection in turns causes a "connection > closed locally" exception to get reported instead of the original exception. > This typically happens if another thread is attempting to read from / write > to the connection concurrently when the original exception that caused the > connection to get closed occurred. > > The fix makes sure that we store the first exception in the underlying > connection before closing it, and that this is the exception that gets > subsequently reported. > > Some minor drive by test fixes. No new regression test. test/jdk/java/net/httpclient/http2/ExpectContinueResetTest.java line 104: > 102: testThrowable = e.getCause(); > 103: } > 104: assertNotNull(exception, "Request should have completed > exceptionally but exception is null"); Unlike the `assertNotNull` on the next line, I think it might be better to replace this new `assertNotNull` with a `fail(...)` inside the try block immediately after the `performRequest()` call. ------------- PR Review Comment: https://git.openjdk.org/jdk/pull/23080#discussion_r1914555193
