/*******************************************************************************
*
* Copyright (c) 2004-2010 Oracle Corporation.
*
* All rights reserved. This program and the accompanying materials
* are made available under the terms of the Eclipse Public License v1.0
* which accompanies this distribution, and is available at
* http://www.eclipse.org/legal/epl-v10.html
*
* Contributors:
*
* InfraDNA, Inc.
*
*
*******************************************************************************/
package hudson;
import com.google.common.collect.ImmutableList;
import hudson.init.InitMilestone;
import net.java.sezpoz.Index;
import net.java.sezpoz.IndexItem;
import hudson.model.Hudson;
import hudson.model.Descriptor;
import java.util.Collections;
import java.util.logging.Logger;
import java.util.logging.Level;
import java.util.List;
import java.util.ArrayList;
import java.util.Collection;
import java.lang.reflect.AnnotatedElement;
import java.lang.reflect.Field;
import java.lang.reflect.Method;
/**
* Discovers the implementations of an extension point.
*
* <p> This extension point allows you to write your implementations of
* {@link ExtensionPoint}s in arbitrary DI containers, and have Hudson discover
* them.
*
* <p> {@link ExtensionFinder} itself is an extension point, but to avoid
* infinite recursion, Hudson discovers {@link ExtensionFinder}s through
* {@link Sezpoz} and that alone.
*
* @author Kohsuke Kawaguchi
* @since 1.286
*/
public abstract class ExtensionFinder implements ExtensionPoint {
/**
* @deprecated as of 1.356 Use and implement {@link #find(Class, Hudson)}
* that allows us to put some metadata.
*/
@Deprecated
public <T> Collection<T> findExtensions(Class<T> type, Hudson hudson) {
return Collections.emptyList();
}
/**
* Discover extensions of the given type.
*
* <p> This method is called only once per the given type after all the
* plugins are loaded, so implementations need not worry about caching.
*
* @param <T> The type of the extension points. This is not bound to
* {@link ExtensionPoint} because of {@link Descriptor}, which by itself
* doesn't implement {@link ExtensionPoint} for a historical reason.
* @param hudson Hudson whose behalf this extension finder is performing
* lookup.
* @return Can be empty but never null.
* @since 1.356 Older implementations provide
* {@link #findExtensions(Class, Hudson)}
*/
public abstract <T> Collection<ExtensionComponent<T>> find(Class<T> type, Hudson hudson);
/**
* A pointless function to work around what appears to be a HotSpot problem.
* See HUDSON-5756 and bug 6933067 on BugParade for more details.
*/
public <T> Collection<ExtensionComponent<T>> _find(Class<T> type, Hudson hudson) {
return find(type, hudson);
}
/**
* Performs class initializations without creating instances.
*
* If two threads try to initialize classes in the opposite order, a dead
* lock will ensue, and we can get into a similar situation with
* {@link ExtensionFinder}s.
*
* <p> That is, one thread can try to list extensions, which results in
* {@link ExtensionFinder} loading and initializing classes. This happens
* inside a context of a lock, so that another thread that tries to list the
* same extensions don't end up creating different extension instances. So
* this activity locks extension list first, then class initialization next.
*
* <p> In the mean time, another thread can load and initialize a class, and
* that initialization can eventually results in listing up extensions, for
* example through static initializer. Such activity locks class
* initialization first, then locks extension list.
*
* <p> This inconsistent locking order results in a dead lock, you see.
*
* <p> So to reduce the likelihood, this method is called in prior to
* {@link #find(Class, Hudson)} invocation, but from outside the lock. The
* implementation is expected to perform all the class initialization
* activities from here.
*
* <p> See http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=6459208 for
* how to force a class initialization. Also see
* http://kohsuke.org/2010/09/01/deadlock-that-you-cant-avoid/ for how class
* initialization can results in a dead lock.
*/
public void scout(Class extensionType, Hudson hudson) {
}
/**
* The default implementation that looks for the {@link Extension} marker.
*
* <p> Uses Sezpoz as the underlying mechanism.
*/
@Extension
public static final class Sezpoz extends ExtensionFinder {
private volatile List<IndexItem<Extension, Object>> indices;
/**
* Loads indices (ideally once but as few times as possible), then reuse
* them later. {@link ExtensionList#ensureLoaded()} guarantees that this
* method won't be called until {@link InitMilestone#PLUGINS_PREPARED}
* is attained, so this method is guaranteed to see all the classes and
* indices.
*/
private List<IndexItem<Extension, Object>> getIndices() {
// this method cannot be synchronized because of a dead lock possibility in the following order of events:
// 1. thread X can start listing indices, locking this object 'SZ'
// 2. thread Y starts loading a class, locking a classloader 'CL'
// 3. thread X needs to load a class, now blocked on CL
// 4. thread Y decides to load extensions, now blocked on SZ.
// 5. dead lock
if (indices == null) {
ClassLoader cl = Hudson.getInstance().getPluginManager().uberClassLoader;
indices = ImmutableList.copyOf(Index.load(Extension.class, Object.class, cl));
}
return indices;
}
public <T> Collection<ExtensionComponent<T>> find(Class<T> type, Hudson hudson) {
List<ExtensionComponent<T>> result = new ArrayList<ExtensionComponent<T>>();
for (IndexItem<Extension, Object> item : getIndices()) {
try {
AnnotatedElement e = item.element();
Class<?> extType;
if (e instanceof Class) {
extType = (Class) e;
} else if (e instanceof Field) {
extType = ((Field) e).getType();
} else if (e instanceof Method) {
extType = ((Method) e).getReturnType();
} else {
throw new AssertionError();
}
if (type.isAssignableFrom(extType)) {
Object instance = item.instance();
if (instance != null) {
result.add(new ExtensionComponent<T>(type.cast(instance), item.annotation()));
}
}
} catch (LinkageError e) {
// sometimes the instantiation fails in an indirect classloading failure,
// which results in a LinkageError
LOGGER.log(item.annotation().optional() ? Level.FINE : Level.WARNING,
"Failed to load " + item.className(), e);
} catch (InstantiationException e) {
LOGGER.log(item.annotation().optional() ? Level.FINE : Level.WARNING,
"Failed to load " + item.className(), e);
}
}
return result;
}
@Override
public void scout(Class extensionType, Hudson hudson) {
for (IndexItem<Extension, Object> item : getIndices()) {
try {
// we might end up having multiple threads concurrently calling into element(),
// but we can't synchronize this --- if we do, the one thread that's supposed to load a class
// can block while other threads wait for the entry into the element call().
// looking at the sezpoz code, it should be safe to do so
AnnotatedElement e = item.element();
Class<?> extType;
if (e instanceof Class) {
extType = (Class) e;
} else if (e instanceof Field) {
extType = ((Field) e).getType();
} else if (e instanceof Method) {
extType = ((Method) e).getReturnType();
} else {
throw new AssertionError();
}
// according to http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=6459208
// this appears to be the only way to force a class initialization
Class.forName(extType.getName(), true, extType.getClassLoader());
} catch (InstantiationException e) {
LOGGER.log(item.annotation().optional() ? Level.FINE : Level.WARNING,
"Failed to scout " + item.className(), e);
} catch (ClassNotFoundException e) {
LOGGER.log(Level.WARNING, "Failed to scout " + item.className(), e);
} catch (LinkageError e) {
LOGGER.log(Level.WARNING, "Failed to scout " + item.className(), e);
}
}
}
}
private static final Logger LOGGER = Logger.getLogger(ExtensionFinder.class.getName());
}