On 29/01/2011 18:48, Konstantin Kolinko wrote:
> Hi!
>
> In r1063366 [1] there was a check added to Nttp11NioProtocol class that
> now causes 3 of 5 tests in TestFormAuthenticator to fail with Nio connector.
That'll be my fault. I'll revisit that patch.
Mark
> I do not know whether you will be able to reproduce them. It somehow depends
> on timing: I am not able to reproduce it when debugging, and out
> buildbot [2] is green:
>
> Anyway, it fails for me reliably when running without debugger.
>
> From TestFormAuthenticator the following three tests are failing:
> testGet
> testPostNoContinue
> testPostWithContinue
>
> The following two tests run successfully:
> testPostNoContinuePostRedirect
> testPostWithContinuePostRedirect
>
> The failure happens when the first read in InternalNioInputBuffer
> returns headers with a single CRLF following them. The test sends the
> second CRLF with some delay.
>
> I wonder why it is not reproducible on the build bot. Different OS? I
> am running with WinXP SP3.
>
> To reproduce:
> 1. Add the following into build.properties:
> execute.test.bio=false
> execute.test.nio=true
> test.entry=org.apache.catalina.authenticator.TestFormAuthenticator
> 2. Add the following to logging.properties:
> org.apache.coyote.level = FINE
> 3. Run "ant test"
>
> The following is in
> logs/TEST-org.apache.catalina.authenticator.TestFormAuthenticator.NIO.txt
> (...)
> 29.01.2011 21:13:46 org.apache.coyote.AbstractProtocolHandler start
> INFO: Starting ProtocolHandler ["http-nio-8001"]
> 29.01.2011 21:13:46
> org.apache.coyote.http11.Http11NioProtocol$Http11ConnectionHandler
> register
> FINE: Register [org.apache.coyote.http11.Http11NioProcessor@c09554] count=1
> 29.01.2011 21:13:46 org.apache.coyote.http11.InternalNioInputBuffer
> parseRequestLine
> FINE: Received [GET /examples/jsp/security/protected/index.jsp HTTP/1.1
> Host: localhost
> Connection: close
> ]
> 29.01.2011 21:13:46
> org.apache.coyote.http11.Http11NioProtocol$Http11ConnectionHandler
> process
> FINE: IOException reading request
> java.io.IOException: Error processing request. Socket is in the long
> poll state but neither Servlet 3.0+ async or Comet is being used
> at
> org.apache.coyote.http11.Http11NioProtocol$Http11ConnectionHandler.process(Http11NioProtocol.java:379)
> at
> org.apache.tomcat.util.net.NioEndpoint$SocketProcessor.run(NioEndpoint.java:1569)
> at
> java.util.concurrent.ThreadPoolExecutor$Worker.runTask(ThreadPoolExecutor.java:886)
> at
> java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:908)
> at java.lang.Thread.run(Thread.java:662)
>
>
> If I remove the check added in r1063366, the test runs successfully:
>
> Index: java/org/apache/coyote/http11/Http11NioProtocol.java
> ===================================================================
> --- java/org/apache/coyote/http11/Http11NioProtocol.java (revision
> 1065026)
> +++ java/org/apache/coyote/http11/Http11NioProtocol.java (working copy)
> @@ -374,10 +374,7 @@
> // complete or dispatch)
> state = processor.asyncPostProcess();
> } else {
> - // Error condition. A connection in this state should
> - // by using one of async or comet
> - throw new IOException(sm.getString(
> - "http11protocol.state.long.error"));
> + socket.getPoller().add(socket);
> }
> }
> if (state == SocketState.LONG || state ==
> SocketState.ASYNC_END) {
>
>
> [1] http://svn.apache.org/viewvc?view=revision&revision=1063366
> [2] http://ci.apache.org/builders/tserver-trunk
>
> Best regards,
> Konstantin Kolinko
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org
> For additional commands, e-mail: dev-h...@tomcat.apache.org
>
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org
For additional commands, e-mail: dev-h...@tomcat.apache.org
