Hi Rory,
I am back at home since a few hours. Will install the new build at some time tomorrow on Jenkins. In the meantime, the MacOS SIGSEGV occurred again - I will try to open a conventional “bugs.sun.com” bug report tomorrow, if there is none already. Uwe ----- Uwe Schindler H.-H.-Meier-Allee 63, D-28213 Bremen <http://www.thetaphi.de/> http://www.thetaphi.de eMail: u...@thetaphi.de From: Rory O'Donnell Oracle, Dublin ireland [mailto:rory.odonn...@oracle.com] Sent: Tuesday, November 05, 2013 11:13 AM To: Uwe Schindler Cc: dev@lucene.apache.org; 'Dalibor Topic'; 'Donald Smith'; 'Seán Coffey'; 'Balchandra Vaidya' Subject: Re: Testing Java Updates with your projects. Hi Uwe, b114 , containing the fix,is now available - https://jdk8.java.net/download.html Very interested to hear your results of testing this build. Rgds,Rory On 10/23/13 02:03 PM, Uwe Schindler wrote: Thanks Rory for the info! I am changing the recipients of this mail, so it no longer goes to the private list. @ dev@lao: FYI, the clone() bug seems to be fixed so we can soon upgrade to JDK8 latest and run tests again. Uwe ----- Uwe Schindler H.-H.-Meier-Allee 63, D-28213 Bremen <http://www.thetaphi.de/> http://www.thetaphi.de eMail: u...@thetaphi.de From: Rory O'Donnell [mailto:rory.odonn...@oracle.com] Sent: Wednesday, October 23, 2013 1:08 PM To: rory.odonn...@oracle.com Cc: Uwe Schindler; 'Dawid Weiss'; 'Robert Muir'; 'Dalibor Topic'; 'Donald Smith'; 'Seán Coffey'; 'Balchandra Vaidya'; priv...@lucene.apache.org Subject: Re: Testing Java Updates with your projects. Hi Uwe, I noticed https://bugs.openjdk.java.net/browse/JDK-8026394 has moved into fixed state. Hope to see this included in the near future, will update you. It is not in b112 <https://jdk8.java.net/download.html> which is not available. Rgds, Rory On 18/10/2013 15:51, Rory O'Donnell Oracle, Dublin Ireland wrote: Hi Uwe, It turns out this is a duplicate of https://bugs.openjdk.java.net/browse/JDK-8026394 Rgds,Rory On 18/10/2013 10:09, Rory O'Donnell Oracle, Dublin Ireland wrote: Hi Uwe, Balchandra has logged a bug for this issue: https://bugs.openjdk.java.net/browse/JDK-8026845 Rgds,Rory On 17/10/2013 18:27, Uwe Schindler wrote: Hi, I was able to reproduce with a simple test case that emulates the UIMA code. See attached test case, just compile it with any JDK and run with b111: With Java 7 or JDK8b109: javac TestCloneInterface.java java TestCloneInterface With JDK8b111: java TestCloneInterface Exception in thread "main" java.lang.IllegalAccessError: tried to access method java.lang.Object.clone()Ljava/lang/Object; from class TestCloneInterface at TestCloneInterface.test(TestCloneInterface.java:15) at TestCloneInterface.main(TestCloneInterface.java:19) The bug happens if the clone() method is declared in a superinterface only. Without the additional interface inbetween, test passes. Instead of the real interface (the "o" local variable, which is of type "FoobarIntf") it checks access flags on "this", which is of type "TestCloneInterface". Uwe ----- Uwe Schindler uschind...@apache.org Apache Lucene PMC Chair / Committer Bremen, Germany http://lucene.apache.org/ -----Original Message----- From: Rory O'Donnell Oracle, Dublin Ireland [mailto:rory.odonn...@oracle.com] Sent: Thursday, October 17, 2013 7:19 PM To: Uwe Schindler Cc: 'Dawid Weiss'; 'Robert Muir'; 'Dalibor Topic'; 'Donald Smith'; 'Seán Coffey'; priv...@lucene.apache.org; Balchandra Vaidya Subject: Re: Testing Java Updates with your projects. Hi Uwe, The more info you can provide the better. Adding Balchandra to email. Rgds,Rory On 17/10/2013 17:41, Uwe Schindler wrote: Hi Rory, we found a new bug in 8 b111, not appearing in Java 7 and Java 8 b109: It seems to be related to one recent commit, reproduces all the time (reproduces with bytecode compiled with JDK8b111 and ran with JDK8b111, but also with bytecode compiled with JDK7 and ran with JDK8b111). I am trying to understand what's happening, but it looks like the patch fails to check the access flags of the method on the correct instance/type/whatever. This is the commit that I think causes this: http://hg.openjdk.java.net/jdk8/jdk8/hotspot/rev/36b97be47bde Issue: https://bugs.openjdk.java.net/browse/JDK-8011311 What happens: When running Apache Lucene's UIMA tests (UIMA is foreign code, not part of Lucene): cd lucene/analysis/uima ant test *All* tests fail here: java.lang.IllegalAccessError: tried to access method java.lang.Object.clone()Ljava/lang/Object; from class org.apache.uima.analysis_engine.impl.AnalysisEngineImplBase at __randomizedtesting.SeedInfo.seed([BC36C2DC5FC6C107:4A94D14D35381F8 8]:0) at org.apache.uima.analysis_engine.impl.AnalysisEngineImplBase.initialize(Anal ysisEngineImplBase.java:163) at org.apache.uima.analysis_engine.impl.AggregateAnalysisEngine_impl.initializ e(AggregateAnalysisEngine_impl.java:127) at org.apache.uima.impl.AnalysisEngineFactory_impl.produceResource(Analysi sEngineFactory_impl.java:94) at org.apache.uima.impl.CompositeResourceFactory_impl.produceResource(C ompositeResourceFactory_impl.java:62) at org.apache.uima.UIMAFramework.produceResource(UIMAFramework.java: 267) at org.apache.uima.UIMAFramework.produceAnalysisEngine(UIMAFramework .java:335) at org.apache.lucene.analysis.uima.ae.BasicAEProvider.getAE(BasicAEProvider.j ava:73) at org.apache.lucene.analysis.uima.BaseUIMATokenizer.analyzeInput(BaseUI MATokenizer.java:63) at org.apache.lucene.analysis.uima.UIMAAnnotationsTokenizer.initializeIterato r(UIMAAnnotationsTokenizer.java:60) at org.apache.lucene.analysis.uima.UIMAAnnotationsTokenizer.incrementToke n(UIMAAnnotationsTokenizer.java:74) at org.apache.lucene.index.DocInverterPerField.processFields(DocInverterPerF ield.java:98) [...] Tests pass with previous versions of JDK8. As this is not our code, it is hard for me to understand (UIMA uses maaaaaaaaaany interfaces...). I just found out: This code fails: http://grepcode.com/file/repo1.maven.org/maven2/org.apache.uima/uimaj - core/2.3.1/org/apache/uima/analysis_engine/impl/AnalysisEngineImplBase .java#163 The interface used here declares clone() correctly and public: http://grepcode.com/file/repo1.maven.org/maven2/org.apache.uima/uimaj - core/2.3.1/org/apache/uima/resource/metadata/MetaDataObject.java#Met aD ataObject.clone%28%29 To me it looks like the access flag check is done on the "this" instance and not on the local variable "md". "md" is definitely not of type org.apache.uima.analysis_engine.impl.AnalysisEngineImplBase like the Exception suggests. The class in the Exception for sure does not implement clone, so it is protected (from Object). It looks like hotspot mixes up the call stack. How should I proceed with opening bug reports? As first step, I think I will now download Apache UIMA and run their tests. This failure is not related to Apache Lucene its just a referenced library in one of our modules. Uwe ----- Uwe Schindler <mailto:uschind...@apache.org> uschind...@apache.org Apache Lucene PMC Chair / Committer Bremen, Germany <http://lucene.apache.org/> http://lucene.apache.org/ -- Rgds, Rory O'Donnell Senior Quality Engineering Manager Java Platform Group Oracle EMEA , Block P5, East Point Business Park, Dublin 3 Phone: +353 (0)1 8033887 -- Rgds, Rory O'Donnell Senior Quality Engineering Manager Java Platform Group Oracle EMEA , Block P5, East Point Business Park, Dublin 3 Phone: +353 (0)1 8033887