* To add tracepoints to a program, you must first decide where to place the * tracepoints, what the logical names are for these points, what information * will be available to the tracing mechanisms at each point, and decide upon * any logical grouping. *
* You add instrumentation to a program in three steps: *
* The method calls representing the tracepoints have no logical * impact on the program. The side effect of the call is that any * activated tracing mechanisms will be notified that the tracepoint has * been hit, and will take whatever actions are appropriate (for example, * logging the tracepoint, or triggering a DTrace probe, etc.). In most * cases, the impact on performance of adding tracepoints to the application * will be minimal. *
* Each logical grouping of tracepoints should be defined in a common * interface, called a provider. An application can have one or many * providers. Each provider is independent and can be created whenever * it is appropriate for that provider, for example, when a subsytem is * initialized. Providers should be disposed of when they are no longer * needed, to free up any associated system resources. Each tracepoint * in a provider is represented by a method in that interface. These methods * are referred to as probes. The method signature determines the probe * parameters. A call to the method with the specified parameters triggers * the probe and makes its parameter values visible to any associated tracing * mechanism. *
* User-defined interfaces which represent providers must extend the * {@code Provider} interface. To activate the system-defined * tracing mechanisms, you must obtain an instance of the * {@code ProviderFactory} class, and pass the class of the provider to * the {@code createProvider()} method. The returned instance is then used to * trigger the probes later in the application. *
* In addition to triggering the probes, the provider instance can be used * to obtain direct references to the {@code Probe} objects, which can be used * directly for triggering, or can be queried to determine whether the probe is * currently being traced. The {@code Provider} interface also defines a * {@code Provider.dispose()} method which is used to free up any resources * that might be associated with that provider. *
* When a probe is triggered, any activated tracing system will be given * the provider name, the probe name, and the values of the probe arguments. * The tracing system is free to consume this data is whatever way is * appropriate. * By default, the provider name is the same as the class name of the interface * that defines the provider. Similarly, the probe name is * the name of the method that defines the probe. These default values * can be over-ridden by annotations. The provider definition can be * annotated with the {@code @ProviderName} annotation, whose value will * indicate the provider name that the tracing system will use. Similarly, * the {@code @ProbeName} annotation annotates a declared method and * indicates the probe name that should be used in the place of the * method name. These annotations can be used to define providers and * probes with the same name, in cases where the semantics of the Java language * may prevent this. *
* Here is a very small and simple usage example: *
*
import com.sun.tracing.Provider;
import com.sun.tracing.ProviderFactory;
interface MyProvider extends Provider {
void startProbe();
void finishProbe(int value);
}
public class MyApplication {
public static void main(String argv[]) {
ProviderFactory factory = ProviderFactory.getDefaultFactory();
MyProvider trace = factory.createProvider(MyProvider.class);
trace.startProbe();
int result = foo();
trace.finishProbe(result);
trace.dispose();
}
}
*
* The Java Development Kit (JDK) currently only includes one system-defined
* tracing framework: DTrace. DTrace is enabled automatically whenever an
* application is run on a system and a JDK release that supports it. When
* DTrace is enabled, probes are made available for listing and matching by
* DTrace scripts as soon as the provider is created. At the tracepoint, an
* associated DTrace script is informed of the creation of the provider, and
* it takes whatever action it is designed to take. Tracepoints in the
* program have the following DTrace probe names:
* The {@code com.sun.tracing.dtrace} package contains additional
* annotations that can be used to control the names used for the
*
* Integer, float and string probe parameters are made available to DTrace
* using
* the built-in argument variables, {@code arg0 ... arg_n}. Integer-types
* are passed by value (boxed values are unboxed), floating-point types are
* passed as encoded integer
* arguments, and {@code java.lang.String} objects are converted
* to UTF8 strings, so they can be read into the DTrace script using the
* {@code copyinstr()} intrinsic. Non-string and non-boxed primitive
* reference arguments are only
* placeholders and have no value.
*
* Using the example above, with a theoretical process ID of 123, these are
* the probes that can be traced from DTrace:
* The DTrace tracing mechanism is enabled for all providers, apart from in the
* following circumstances:
*
*/
package com.sun.tracing;
* {@code
*
* module and function fields, as well as annotations
* that can be added to the provider to control probe stability and dependency
* attributes.
*
MyProvider123:::startProbe
MyProvider123:::finishProbe
* When {@code finishProbe} executes, {@code arg0} will contain the
* value of {@code result}.
*
*
*