On 17.12.2013 11:01, Rainer Jung wrote: > A) TC 7 > ======= > > 1) org.apache.catalina.loader.TestWebappClassLoaderMemoryLeak > ------------------------------------------------------------- > > 1 x apr, 3 x bio, 4 x nio
Fixed by syncing the test impl with trunk which uses a wait loop. No more failures observed. > 2) org.apache.catalina.websocket.TestWebSocket > ---------------------------------------------- > > 2 x bio, 1 x nio I think I know what the problem is and I'm just now running an ugly fix in a loop. Looks good. Problem: the web socket client in the test class reads from the socket input stream first using a buffered reader (for the headers, encoding ISO-8859-1) and then later using the raw input stream (for the body, decoding using a B2CConverter to UTF-8). Sometimes the body can not be found in the raw stream, because it already was slurped in by the buffered reader. Unfortunately simply using an InputStreamReader instead of a BufferedReader doesn't fix the problem, because the InputStreamReader still seems to buffer a few bytes sometimes. So I'm now doing I/O only on the InputStream, converting to ISO-8859-1 resp. UTF-8 using B2CConverter. It seems to work, but is a bit more code and also a bit ugly. I found no clean way of switching the decoding charset with JDK APIs though. So if I'm right those test failures were false positives. Regards, Rainer --------------------------------------------------------------------- To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org For additional commands, e-mail: dev-h...@tomcat.apache.org