MethodHandleBuilder bug? (original) (raw)

Paul Sandoz paul.sandoz at oracle.com
Thu Apr 12 21:25:40 UTC 2018

Previous message (by thread): MethodHandleBuilder bug?
Next message (by thread): MethodHandleBuilder bug?
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

Try with the patch below.

I ain’t tested this throughly.

Paul.

diff -r 4d1e8489a000 src/java.base/share/classes/jdk/experimental/bytecode/BasicTypeHelper.java --- a/src/java.base/share/classes/jdk/experimental/bytecode/BasicTypeHelper.java Thu Apr 12 09:26:53 2018 -0400 +++ b/src/java.base/share/classes/jdk/experimental/bytecode/BasicTypeHelper.java Thu Apr 12 14:22:51 2018 -0700 @@ -73,6 +73,9 @@ case 'V': return TypeTag.V; default:

```
           if (s == nullType()) {
```
```
               return TypeTag.A;
```
```
           }
           throw new IllegalStateException("Bad type: " + s);
   }
```
} diff -r 4d1e8489a000 src/java.base/share/classes/jdk/experimental/bytecode/TypedCodeBuilder.java --- a/src/java.base/share/classes/jdk/experimental/bytecode/TypedCodeBuilder.java Thu Apr 12 09:26:53 2018 -0400 +++ b/src/java.base/share/classes/jdk/experimental/bytecode/TypedCodeBuilder.java Thu Apr 12 14:22:51 2018 -0700 @@ -238,6 +238,7 @@ State(ArrayList stack, Vector locals) { this.stack = stack; this.locals = locals;

       this.alive = true;
   }

   State() {

On Apr 12, 2018, at 1:23 PM, Paul Sandoz <paul.sandoz at oracle.com> wrote:

Thanks, i think there is a bug processing the label when merging stack state from which to generate a stack map entry. Something to do with liveness… i might be able to work it out more fully... but i am not too familiar with stack map entries. Paul.

On Apr 12, 2018, at 7:37 AM, Roland Westrelin <rwestrel at redhat.com> wrote:

I noticed that this: private static final MethodHandle mergeNull = MethodHandleBuilder.loadCode(MethodHandles.lookup(), "mergeNull", MethodType.methodType(void.class, TestLWorld.class, boolean.class), CODE -> { CODE. iload1(). iconst0(). ifcmp(TypeTag.I, CondKind.EQ, "null"). aload0(). getfield(TestLWorld.class, "valueField1", "Lcompiler/valhalla/valuetypes/MyValue1;"). goto("continue"). label("null"). aconstnull(). label("continue"). aload0(). swap(). putfield(TestLWorld.class, "valueField1", "Lcompiler/valhalla/valuetypes/MyValue1;"). return(); } ); works fine but if I swap the code in the if branches: private static final MethodHandle mergeNull = MethodHandleBuilder.loadCode(MethodHandles.lookup(), "mergeNull", MethodType.methodType(void.class, TestLWorld.class, boolean.class), CODE -> { CODE. iload1(). iconst0(). ifcmp(TypeTag.I, CondKind.EQ, "null"). aconstnull(). goto("continue"). label("null"). aload0(). getfield(TestLWorld.class, "valueField1", "Lcompiler/valhalla/valuetypes/MyValue1;"). label("continue"). aload0(). swap(). putfield(TestLWorld.class, "valueField1", "Lcompiler/valhalla/valuetypes/MyValue1;"). return(); } ); then I get: java.lang.ExceptionInInitializerError at java.base/jdk.internal.reflect.NativeMethodAccessorImpl.invoke0(Native Method) at java.base/jdk.internal.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:62) at java.base/jdk.internal.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43) at java.base/java.lang.reflect.Method.invoke(Method.java:564) at com.sun.javatest.regtest.agent.MainWrapper$MainThread.run(MainWrapper.java:115) at java.base/java.lang.Thread.run(Thread.java:841) Caused by: java.lang.IllegalStateException: Bad type: at java.base/jdk.experimental.bytecode.BasicTypeHelper.tag(BasicTypeHelper.java:76) at java.base/jdk.experimental.value.MethodHandleBuilder$IsolatedMethodBuilder$IsolatedMethodTypeHelper.tag(MethodHandleBuilder.java:196) at java.base/jdk.experimental.value.MethodHandleBuilder$IsolatedMethodBuilder$IsolatedMethodTypeHelper.tag(MethodHandleBuilder.java:138) at java.base/jdk.experimental.bytecode.CodeBuilder.writeStackMapType(CodeBuilder.java:1257) at java.base/jdk.experimental.bytecode.CodeBuilder.sameLocals1StackItemFrame(CodeBuilder.java:1305) at java.base/jdk.experimental.bytecode.TypedCodeBuilder.writeStackMapFrame(TypedCodeBuilder.java:1231) at java.base/jdk.experimental.bytecode.TypedCodeBuilder.emitStackMap(TypedCodeBuilder.java:1196) at java.base/jdk.experimental.bytecode.TypedCodeBuilder.resolveJumps(TypedCodeBuilder.java:1189) at java.base/jdk.experimental.bytecode.MacroCodeBuilder.label(MacroCodeBuilder.java:437) at compiler.valhalla.valuetypes.TestLWorld.lambda$static$3(TestLWorld.java:505) at java.base/jdk.experimental.bytecode.MethodBuilder.withCode(MethodBuilder.java:50) at java.base/jdk.experimental.value.MethodHandleBuilder.lambda$loadCode$1(MethodHandleBuilder.java:101) at java.base/jdk.experimental.bytecode.ClassBuilder.withMethod(ClassBuilder.java:204) at java.base/jdk.experimental.value.MethodHandleBuilder.loadCode(MethodHandleBuilder.java:99) at java.base/jdk.experimental.value.MethodHandleBuilder.loadCode(MethodHandleBuilder.java:77) at java.base/jdk.experimental.value.MethodHandleBuilder.loadCode(MethodHandleBuilder.java:72) at compiler.valhalla.valuetypes.TestLWorld.(TestLWorld.java:492) ... 6 more Roland.

Previous message (by thread): MethodHandleBuilder bug?
Next message (by thread): MethodHandleBuilder bug?
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

More information about the valhalla-dev mailing list