Experimental Physics and
Industrial Control System

"Kasemir, Kay" <[email protected]> · Tue, 31 Mar 2009 16:27:32 -0400

Hi:

I think JCA 2.3.1 and also the latest 2.3.2 have a small memory
leak in JNIChannel.addMonitor() resp. JNIMonitor.clear():

They ultimately call JNI code in JNI.cpp, where
"new MonitorID" allocates a small structure that's not
freed up in the clearMonitor counterpart:

JNIEXPORT jlong JNICALL Java_gov_aps_jca_jni_JNI__1ch_1addMonitor(JNIEnv
*env, jclass, jint type, jint length, jlong channelID, jobject callback,
jint mask)
{
  MonitorID* monitorID= new MonitorID;
  monitorID->callback= env->NewGlobalRef(callback);
  checkstatus( env, ca_add_masked_array_event(type,length,CHID(channelID),
&theArrayMonitorCallback,
monitorID->callback,0,0,0,&monitorID->pevid,mask));
  return (jlong)monitorID;
}

JNIEXPORT void JNICALL Java_gov_aps_jca_jni_JNI__1ch_1clearMonitor(JNIEnv
*env, jclass, jlong monitorID)
{
  checkstatus( env, ca_clear_event(((MonitorID*)monitorID)->pevid) );
  env->DeleteGlobalRef(((MonitorID*)monitorID)->callback);

 // Needs something like this?
 // delete ((MonitorID*)monitorID)
}

Thanks,
-Kay

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System