Incomplete implementation of GetFieldID and GetStaticFieldID

David Grove added a comment - 17/Jun/09 5:02 PM The patch is more complex than it needs to be to fix the problem. Working on simpler fix.

David Grove added a comment - 17/Jun/09 9:30 PM fix and test cases committed in r15712.

Byeong Lee added a comment - 18/Jun/09 10:47 PM - edited The r15712 does not yet complete the GetStaticFieldID, and here is an updated testcase. m$svn diff testing/tests/jni/src/TestJNIGetFieldID.java Index: testing/tests/jni/src/TestJNIGetFieldID.java =================================================================== --- testing/tests/jni/src/TestJNIGetFieldID.java (revision 15712) +++ testing/tests/jni/src/TestJNIGetFieldID.java (working copy) @@ -74,6 +74,18 @@ allTestPass = false; } + try { + if (getStaticFinalF(I.class) == 1) { + if (verbose) System.out.println("static_I.f: pass"); + } else { + if (verbose) System.out.println("static_I.f: fail"); + allTestPass = false; + } + } catch(Throwable e) { + if (verbose) System.out.println("static_f: fail"); + allTestPass = false; + } + if (allTestPass) { System.out.println("PASS: TestJNIGetFieldID"); } else { Here is a failure in Jikes RVM, and successes in Hotspot and J9: $rvm -showversion -Djava.library.path=./target/tests/byeong/prototype/jni -classpath ./target/tests/byeong/prototype/jni/classes TestJNIGetFieldID production Jikes RVM 3.1.0+svn (r15711M) instance_a: pass static_s: pass static_f: pass static_f: fail FAIL: TestJNIGetFieldID $java -showversion -Djava.library.path=./target/tests/byeong/prototype/jni -classpath ./target/tests/byeong/prototype/jni/classes TestJNIGetFieldID java version "1.6.0_13" Java(TM) SE Runtime Environment (build 1.6.0_13-b03) Java HotSpot(TM) Server VM (build 11.3-b02, mixed mode) instance_a: pass static_s: pass static_f: pass static_I.f: pass PASS: TestJNIGetFieldID $java -showversion -Djava.library.path=./target/tests/byeong/prototype/jni -classpath ./target/tests/byeong/prototype/jni/classes TestJNIGetFieldID java version "1.6.0" Java(TM) SE Runtime Environment (build pxi3260sr4-20090219_01(SR4)) IBM J9 VM (build 2.4, J2RE 1.6.0 IBM J9 2.4 Linux x86-32 jvmxi3260-20090215_29883 (JIT enabled, AOT enabled) J9VM - 20090215_029883_lHdSMr JIT - r9_20090213_2028 GC - 20090213_AA) JCL - 20090218_01 instance_a: pass static_s: pass static_f: pass static_I.f: pass PASS: TestJNIGetFieldID Here is a proposed patch and the result: $svn diff rvm/src/org/jikesrvm/jni/JNIFunctions.java Index: rvm/src/org/jikesrvm/jni/JNIFunctions.java =================================================================== --- rvm/src/org/jikesrvm/jni/JNIFunctions.java (revision 15712) +++ rvm/src/org/jikesrvm/jni/JNIFunctions.java (working copy) @@ -3447,12 +3447,13 @@ } } } - // Now search all implemented interfaces (includes inherited interfaces) - for (RVMClass curClass : rvmType.asClass().getAllImplementedInterfaces()) { - for (RVMField field : curClass.getStaticFields()) { - if (field.getName() == fieldName && field.getDescriptor() == descriptor) { + } + + // Now search all implemented interfaces (includes inherited interfaces) + for (RVMClass curClass : rvmType.asClass().getAllImplementedInterfaces()) { + for (RVMField field : curClass.getStaticFields()) { + if (field.getName() == fieldName && field.getDescriptor() == descriptor) { return field.getId(); - } } } } $rvm -showversion -Djava.library.path=./target/tests/byeong/prototype/jni -classpath ./target/tests/byeong/prototype/jni/classes TestJNIGetFieldID prototype Jikes RVM 3.1.0+svn (r15712M) instance_a: pass static_s: pass static_f: pass static_I.f: pass PASS: TestJNIGetFieldID The RVM r15711rejects an interface class, which the JNI specification allows: ". The field may be defined in the class or interface referred to by clazz or in one of its superclass or superinterfaces. " http://java.sun.com/docs/books/jni/html/functions.html#65115

David Grove added a comment - 19/Jun/09 10:23 AM Hi BK, Are you actually working with a more complex example in which the order of traversing superclasses vs. superinterfaces matters? The current svn head passes the extended test case for me. So unless it is a question of ordering (the field is declared in both an interface of this class and as a field of a superclass and we are supposed to find the interface one in preference to the superclass one), then I don't see why it isn't working for you.

David Grove added a comment - 19/Jun/09 11:22 AM ok, Appreciate getting nice test cases along with the bug report...thanks!