Hi Volker, I fully support this change, although I'm not qualified to review it as I'm not so acquainted with VM internals. The code looks reasonable to me though. This change will greatly help track down class initializiation bugs. I don't think that attaching a cause which is not directly an exception caught at the place where new exception is thrown will confuse people much, although I can imagine that they are used to observe such chains most of the time. The stack traces of cause and top exception clearly show where their source is and as you say the type of cause (ExceptionInInitializerError) hints at that too. Another hint that you may consider adding or not, depending on whether it would be possible to implement elegantly, is for the cause to somehow include the name of the thread in which it occurred in its message. When the top exception (NoClassDefFoundError) is caught and logged, the logger usually includes the thread name. If the cause also included thread name and it was different, it would be another hint to the user of what actually happened. Just some comments / questions about the test. This is really quite a complicated play of exceptions, causes and initialization orders, I must say. I'll try to evaluate my reasoning by describing what test does, so let's start with this: 109 public static class A { 110 public static final int i = (new Object[1])[2].hashCode(); 111 } 112 public static class B extends A {} 113 public static class C extends B {} Those 3 classes are designed so that their initialization fails. Initialization of A fails because ArrayIndexOutOfBoundsException is thrown in field initializer, initialization of B fails because it is a subclass of A and initialization of C fails because it is a subclass of B. On top of that, the test includes the following class: 115 static class ClassWithFailedInitializer { 116 public static final int i; 117 static { 118 cl = new URLClassLoader(new URL[] { 119 ClassWithFailedInitializer.class.getProtectionDomain().getCodeSource().getLocation() }, null); 120 try { 121 cl.loadClass("NoClassDefFoundErrorTest$C").newInstance(); 122 } catch (ClassNotFoundException | InstantiationException | IllegalAccessException e) { 123 throw new RuntimeException("Class should be found", e); 124 } 125 i = 42; 126 } 127 } ...and pulls the trigger with: 137 if (ClassWithFailedInitializer.i == 0) { This triggers initialization of ClassWithFailedInitializer 1st, which constructs new ClassLoader which it then uses to load class C and attempts to instantiate an object from it. You then have the following handler arround the last two actions: 120 try { 121 cl.loadClass("NoClassDefFoundErrorTest$C").newInstance(); 122 } catch (ClassNotFoundException | InstantiationException | IllegalAccessException e) { 123 throw new RuntimeException("Class should be found", e); 124 } ClassNotFoundException should not be thrown as C class should be loadable. That's OK. The other two exceptions that you catch here are a little confusing, because they can only occur after successful class C initialization and because they are impossible exceptions in the concrete setup. But I can understand that you have to handle them somehow because they are checked exceptions. So perhaps you could group them into another catch block and throw RuntimeException with a more descriptive message (like "Impossible situation" or such). Just to be less confusing for someone who would have to read the code in the future... Then we get to the following confusing point: 136 try { 137 if (ClassWithFailedInitializer.i == 0) { 138 throw new RuntimeException("Class initialization succeeded but is designed to fail."); 139 } 140 throw new Exception("Expected exception was not thrown."); 141 } 142 catch (ExceptionInInitializerError e) { 143 e.printStackTrace(); 144 Asserts.assertNE(e.getCause(), null, "Expecting cause in ExceptionInInitializerError."); 145 Asserts.assertEQ(e.getCause().getClass(), ArrayIndexOutOfBoundsException.class, "sanity"); 146 } I think that line 138 is not reachable. If ClassWithFailedInitializer.i == 0 evaluated successfully, it would mean that ClassWithFailedInitializer initialization finished normally, but if it finished normally, then field i would contain 42, wouldn't it? I think you wanted the test to be if (ClassWithFailedInitializer.i == 42) and then you don't need line 140. Am I right? The same goes for line 156. Good work. Regards, Peter On 06/29/18 16:53, Volker Simonis wrote:
Hi,
can I please have a review for the following change which saves ExceptionInInitializerError thrown during class initialization and chains them as cause into potential NoClassDefFoundErrors for the same class. We are using this features since years in our commercial SAP JVM and it proved extremely useful for detecting and fixing errors especially in big deployments.
This is a follow-up on a discussion previously started by Goetz [1]. His first proposal (which is close to our current, internal implementation) inserted an additional field into java.lang.Class objects to save potential ExceptionInInitializerErrors. This was considered to much overhead in the initial discussion [1].
http://cr.openjdk.java.net/~simonis/webrevs/2018/8203826.v2/ https://bugs.openjdk.java.net/browse/JDK-8203826
So in this change, I've completely re-implemented the feature by using a java.lang.Hashtable which is attached to the ClassLoaderData object. The Hashtable is lazily created when the first ExceptionInInitializerError is thrown and maps the Class which triggered the ExceptionInInitializerError during the execution of its static initializer to the corresponding ExceptionInInitializerError.
If the same class will be accessed once again, this will directly lead to a plain NoClassDefFoundError (as per the JVMS, 5.5 Initialization) because the static initializer won't be executed a second time. Until now, this NoClassDefFoundError wasn't linked in any way to the root cause of the problem (i.e. the first ExceptionInInitializerError together with the chained exception that happened during the execution of the static initializer). With this change, the NoClassDefFoundError will now chain the initial ExceptionInInitializerError as cause, making it much easier to detect the problem which lead to the NoClassDefFoundError.
Following is an example from the new JTreg tests which comes which this change to demonstrate the feature. Until know, a typical stack trace from a NoClassDefFoundError looked as follows:
java.lang.NoClassDefFoundError: Could not initialize class NoClassDefFound$ClassWithFailedInitializer at java.base/java.lang.Class.forName0(Native Method) at java.base/java.lang.Class.forName(Class.java:291) at NoClassDefFound.main(NoClassDefFound.java:38)
With this change, the same stack trace now looks as follows:
java.lang.NoClassDefFoundError: Could not initialize class NoClassDefFound$ClassWithFailedInitializer at java.base/java.lang.Class.forName0(Native Method) at java.base/java.lang.Class.forName(Class.java:315) at NoClassDefFound.main(NoClassDefFound.java:38) Caused by: java.lang.ExceptionInInitializerError at java.base/jdk.internal.reflect.NativeConstructorAccessorImpl.newInstance0(Native Method) at java.base/jdk.internal.reflect.NativeConstructorAccessorImpl.newInstance(NativeConstructorAccessorImpl.java:62) at java.base/jdk.internal.reflect.DelegatingConstructorAccessorImpl.newInstance(DelegatingConstructorAccessorImpl.java:45) at java.base/java.lang.reflect.Constructor.newInstance(Constructor.java:490) at java.base/java.lang.Class.newInstance(Class.java:584) at NoClassDefFound$ClassWithFailedInitializer.<clinit>(NoClassDefFound.java:20) at java.base/java.lang.Class.forName0(Native Method) at java.base/java.lang.Class.forName(Class.java:315) at NoClassDefFound.main(NoClassDefFound.java:30) Caused by: java.lang.ArrayIndexOutOfBoundsException: Index 2 out of bounds for length 1 at NoClassDefFound$A.<clinit>(NoClassDefFound.java:9) ... 9 more
As you can see, the reason for the NoClassDefFoundError when accessing the class 'NoClassDefFound$ClassWithFailedInitializer' is actually not even in the class or its static initializer itself, but in the class 'NoClassDefFound$A' which is a base class of 'NoClassDefFound$ClassWithFailedInitializer'. This is not easily detectible from the old, plain NoClassDefFoundError.
As I wrote, the only overhead we have with the new implementation is an additional OopHandle field per ClassLoaderData which I think is acceptable. The Hashtable object itself is only created lazily, after the first occurrence of an ExceptionInInitializerError in the corresponding class loader. The whole Hashtable creation and storing/quering of ExceptionInInitializerErrors in ClassLoaderData::record_init_exception()/ClassLoaderData::query_init_exception() is optional in the sense that any errors/exceptions occurring during the execution of these functions are ignored and cleared.
Finally, we also take care to recursively convert all native backtraces in the stored ExceptionInInitializerErrors (and their suppressed and chained exceptions) into symbolic stack traces in order to avoid holding references to classes and prevent their unloading. This is implemented in the new private, static method java.lang.Throwable::removeNativeBacktrace() which is called for each ExceptionInInitializerError before it is stored in the Hashtable.
Thank you and best regards, Volker
[1] http://mail.openjdk.java.net/pipermail/hotspot-runtime-dev/2018-June/028310....