Experimental Physics and
Industrial Control System

Hi Marty,

 

Thanks for your reply. In the real application, I don’t have any control over the threads (the calls to Get, RPC are done through a GUI that makes calls to a server which does the request). Though I’ve created simpler test applications that both include threading, and don’t, and the same issue occurs.

 

I’ve created a test application that demonstrates the issue (see attached)

 

There are two files – PVATest.java and EpicsTestServer.java.

 

The EpicsTestServer.java is just a simple server that listens for connections, and has one pv – “state”. When it receives an RPC call, it changes the state and just pauses for 5 seconds. This can be started as a standard Java application using the main() method, and should be started before running the tests.

 

The PVATest.java is a JUnit test file with two tests in. The main test that demonstrates the issue is called TestSimulatenous(). It makes an asynchronous RPC call, and then immediately after does a GET. The test fails because the “Status status2 = pvaChannel2.waitConnect(3);” call in the GET section can’t connect with error ‘channel not connected’

 

The test can be made to pass by uncommenting the “//Thread.sleep(5000);” which gives the server time to finish processing the RPC call and return.

 

The other test in this class is called TestSimulatenousOnlyGet() and is a copy of the GET section of the first test. If it is run at the same time as the first (i.e. whilst the server is still pausing during processing of the RPC request) then it can connect and get the state value, showing that the issue isn’t in the server. To get the timings right on this test, I normally debug the tests and the Test Server, with debug points on the line after “// Do Get” in the first test, and on the “Thread.sleep(5000);” line in the Test Server. This ensures the server is still processing the request when the second test makes its GET request.

 

Thanks again for your help,

 

Matt