On 21.09.2011 12:44, Mark Thomas wrote: > On 21/09/2011 01:41, Konstantin Kolinko wrote: >> 2011/9/20 Mark Thomas <ma...@apache.org>: >>> On 20/09/2011 13:41, Konstantin Kolinko wrote: >>>> >>>> BTW: I am unable to reproduce Gump failure for TestCometProcessor with NIO >>>> (running with JDK 6u26/WinXP), >>>> but this test fails for me when running it with APR connector. >>> >>> I *think* I have all these issues fixed now. The unit tests pass for me >>> on Linux and Windows. We'll find out tomorrow if Gump is happy now. If >>> you could test APR as well that would be great. > > Thanks for the test results. > >> Running trunk @r1173433, >> BIO - OK (it actually skips the test) >> APR - OK >> NIO failed first 2 of 8 runs: >> >> [[[ >> ------------- ---------------- --------------- >> >> Testcase: testSimpleCometClient took 17,813 sec >> FAILED >> expected:<Client: READ: [4] bytes> but was:<Client: READ: [8] bytes> >> junit.framework.AssertionFailedError: expected:<Client: READ: [4] >> bytes> but was:<Client: READ: [8] bytes> >> at >> org.apache.catalina.comet.TestCometProcessor.testSimpleCometClient(TestCometProcessor.java:97) >> >> Testcase: testCometConnectorStop took 5,328 sec >> ]]] >> >> IIRC it is expected (two subsequent network packets arriving to Tomcat >> without a delay between them) > > That is unusual but not unexpected. We could handle that to prevent > false positives if you find it happening a lot.
That's the type of failure I saw intermittent with the old code. I didn't yet have a chance to retest. Last time I saw those failures it was always the first two "PING" messages that came in combined as PINGPING. When I added a 2 second delay before starting the write in the PingWriterThread, I could no longer reproduce the problem. Regards, Rainer --------------------------------------------------------------------- To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org For additional commands, e-mail: dev-h...@tomcat.apache.org
