Dan, Meanwhile, I've tested with 2.2.6 and 2.2.7-SNAPSHOT, neither made any difference. See http://old.nabble.com/file/p27895402/cxf-concurrency-issue.txt cxf-concurrency-issue.txt for stack traces produced by 2.2.7-SNAPSHOT, and a list of server side jars in WEB-INF/lib at the time of execution. The web app was running in Tomcat 6.0.24 outside Eclipse, while Test units were run inside Eclipse 3.5. Here's a list of the jars in the project's WEB-INF/lib: http://old.nabble.com/file/p27895402/cxf-concurrency-issue-eclipse-jars.txt cxf-concurrency-issue-eclipse-jars.txt .
Java was JDK 1.6.0.17-b04, client & server on the same machine (XP SP3) with an Intel Dual Core. A typical test run creates 50 threads each doing 10 logins and logouts. Some runs are perfekt, some produce one or both of the described exceptions (mostly only one or the other). Repeated tests with SoapUI worked pretty well, even with considerably higher load than with our unit tests. So the problem seems to lie with the client? Also, logins and logouts were all paired according to the server logs. The client logs missed exactly as many logouts as there were exceptions thrown, in our case meaning that the prior login request had failed. I've downloaded Woodstox 4.0.7. Will placing it on the classpath be enough to make CXF use it, and will I need the MVS jars to? I put a cxf.xml into WEB-INF/classes to log HTTP traffic using the second sample from the online documentation. Can I make CXF log into a file instead of onto the console? If I make it use log4j as described in various places, where do I have to put the log4j.xml / log4j.properties? Would you suggest trying contract first clients? Thanks a lot, Michael dkulp wrote: > > > Any chance you could test this with 2.2.6 or even the 2.2.7-SNAPSHOT? > Any > required fix would be put there, however, I'm really more interested in > the > line numbers for the stack traces with recent code. Easier to map in and > figure out what's happening. > > Also, do you know if woodstox is being used for the parser? If not, can > you > try that? > > Dan > > -- View this message in context: http://old.nabble.com/Concurrency-issue-tp27882337p27895402.html Sent from the cxf-user mailing list archive at Nabble.com.
