Repeated calls to rosjava parameter server produce unexpected results.
During repeated runs with a rosjava master mixing java and C++ nodes (all on Windows, compiled with MSVC), it appeared that a socket connection was dying within the Java code. After some period of time one or more of the C++ nodes would appear to block somewhere down in OS code. Was unable to debug effectively. However, it was noted that the C++ nodes affected were making multiple default parameter calls at 1HZ to the rosjava master:
nh.param<std::string>("default_param", default_param, "default_value");
After removing these calls, the nodes worked reliably. It seems like repeated calls to the java implementation of the parameter server was causing some kind of hang up. Anyone experienced this? I'll try to debug more and get a stack trace as time allows.