michael-o commented on a change in pull request #65:
URL: https://github.com/apache/maven-resolver/pull/65#discussion_r456885102
##########
File path:
maven-resolver-impl/src/main/java/org/eclipse/aether/internal/impl/DefaultSyncContextFactory.java
##########
@@ -28,31 +30,59 @@
import org.eclipse.aether.artifact.Artifact;
import org.eclipse.aether.impl.SyncContextFactory;
import org.eclipse.aether.metadata.Metadata;
+import org.slf4j.Logger;
+import org.slf4j.LoggerFactory;
+
+import com.google.inject.Singleton;
/**
- * A factory to create synchronization contexts. This default implementation
actually does not provide any real
- * synchronization but merely completes the repository system.
+ * A factory to create synchronization contexts. This default implementation
uses fair global locking
+ * based on {@link ReentrantReadWriteLock}. Explicit artifacts and metadata
passed are ignored.
*/
@Named
+@Singleton
public class DefaultSyncContextFactory
implements SyncContextFactory
{
+ private final ReentrantReadWriteLock lock = new ReentrantReadWriteLock(
true );
public SyncContext newInstance( RepositorySystemSession session, boolean
shared )
{
- return new DefaultSyncContext();
+ return new DefaultSyncContext( shared ? lock.readLock() :
lock.writeLock(), shared );
}
static class DefaultSyncContext
implements SyncContext
{
+ private static final Logger LOGGER = LoggerFactory.getLogger(
DefaultSyncContext.class );
+
+ private final Lock lock;
+ private final boolean shared;
+ private int lockHoldCount;
+
+ DefaultSyncContext( Lock lock, boolean shared )
+ {
+ this.lock = lock;
+ this.shared = shared;
+ }
public void acquire( Collection<? extends Artifact> artifact,
Collection<? extends Metadata> metadata )
{
+ LOGGER.trace( "Acquiring global {} lock (currently held: {})",
+ shared ? "read" : "write", lockHoldCount );
+ lock.lock();
Review comment:
@Tibor17 Please read the API of `SyncContextFactory`. Sync contexts are
never shared, they are thread private and everything runs in try-with-resources
clauses. This is the contract. If you don't stick to do, the behavior is
undefined. The current code as well as the change stick to it. The count is
necessary to release reentrant acquired locks. I don't see how a
`CountDownLatch` would help here. I agree with @cstamas . It serves a
completely different purpose. I read the `SyncContextFactory` documentation at
least five times to understand what I expected from an implementation.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
