[svn.haxx.se] · SVN Dev · SVN Users · SVN Org · TSVN Dev · TSVN Users · Subclipse Dev · Subclipse Users · this month's index

Re: [PATCH] Fix for a JNI bug in CreateJ.cpp (JavaHL)

From: Greg Stein <gstein_at_gmail.com>
Date: Wed, 19 May 2010 08:57:26 -0400

Cool. Thanks!!

On Wed, May 19, 2010 at 00:21, Byeongcheol Lee <lineonking_at_gmail.com> wrote:
> [[[
> * subversion/bindings/javahl/native/CreateJ.cpp
>  (Set): Use CallBooleanMethod for a method returning a boolean value.
> ]]]
>
> On Tue, May 18, 2010 at 10:25 PM, Hyrum K. Wright
> <hyrum_wright_at_mail.utexas.edu> wrote:
>>
>>
>> On Tue, May 18, 2010 at 10:14 PM, Byeongcheol Lee <lineonking_at_gmail.com>
>> wrote:
>>>
>>> Dear Subversion Developers:
>>>
>>> I attach a patch for a JNI bug. To reproduce the bug, run your JavaHL
>>> regression test under our dynamic JNI bug detector, Jinn
>>> [http://userweb.cs.utexas.edu/~bclee/jinn].
>>>
>>> $env JAVA_TOOL_OPTIONS=-agentlib:jinn make check-javahl
>>> ...
>>> .Exception in thread "main" xtc.lang.blink.agent.JNIAssertionFailure:
>>> The return type mismatch in CallObjectMethodV: method 0x9626fe4 has
>>> return type: (Ljava/lang/Object;)Z
>>>        at
>>> xtc.lang.blink.agent.JNIAssertionFailure.assertFail(JNIAssertionFailure.java:16)
>>>        at org.apache.subversion.javahl.SVNClient.doImport(Native Method)
>>>        at org.apache.subversion.javahl.SVNTests.setUp(SVNTests.java:239)
>>>        at junit.framework.TestCase.runBare(TestCase.java:128)
>>>        at junit.framework.TestResult$1.protect(TestResult.java:106)
>>>        at junit.framework.TestResult.runProtected(TestResult.java:124)
>>>        at junit.framework.TestResult.run(TestResult.java:109)
>>>        at junit.framework.TestCase.run(TestCase.java:120)
>>>        at junit.framework.TestSuite.runTest(TestSuite.java:230)
>>>        at junit.framework.TestSuite.run(TestSuite.java:225)
>>>        at junit.framework.TestSuite.runTest(TestSuite.java:230)
>>>        at junit.framework.TestSuite.run(TestSuite.java:225)
>>>        at junit.textui.TestRunner.doRun(TestRunner.java:121)
>>>        at junit.textui.TestRunner.doRun(TestRunner.java:114)
>>>        at junit.textui.TestRunner.run(TestRunner.java:77)
>>>        at org.apache.subversion.javahl.RunTests.main(RunTests.java:116)
>>> ...
>>>
>>> To understand what caused the JNI error, I attached a gdb to the JVM.
>>> Relevant caling context and source lines are here.
>>>
>>> #5  0x8f86cdf5 in CreateJ::Set (objects=...)
>>>    at subversion/bindings/javahl/native/CreateJ.cpp:912
>>> #6  0x8f86bf4f in CommitMessage::getCommitMessage (this=0x81bf6d0,
>>>    commit_items=0x817ea60)
>>>    at subversion/bindings/javahl/native/CommitMessage.cpp:210
>>> ...
>>> #11 0x8f88bff0 in Java_org_apache_subversion_javahl_SVNClient_doImport (
>>>    env=0x805a518, jthis=0xb7281ce4, jpath=0xb7281ce0, jurl=0xb7281cdc,
>>>    jmessage=0x0, jdepth=0xb7281cd4, jnoIgnore=0 '\000',
>>>    jignoreUnknownNodeTypes=0 '\000', jrevpropTable=0x0)
>>>    at
>>> subversion/bindings/javahl/native/org_apache_subversion_javahl_SVNClient.cpp:792
>>>
>>> In subversion/bindings/javahl/native/CreateJ.cpp
>>> ...
>>>   883    jclass clazz = env->FindClass("java/util/HashSet");
>>> ...
>>>   895    static jmethodID add_mid = 0;
>>> ...
>>>   898        add_mid = env->GetMethodID(clazz, "add",
>>> "(Ljava/lang/Object;)Z");
>>> ...
>>>   912        env->CallObjectMethod(set, add_mid, jthing);
>>> ...
>>>
>>>
>>> There is a JNI type mismatch between Line 898 and 912.
>>>  The "add_mid" at Line 898 points a Java method returning a boolean
>>> value, but the
>>>  JNI function at Line 912 assumes that "add_mid" is returning a Java
>>> reference.
>>> This violates a JNI programming rule:
>>>
>>> "You should replace type in Call<type>Method with the Java type of the
>>> method you are calling"
>>>
>>>  [http://java.sun.com/javase/6/docs/technotes/guides/jni/spec/functions.html#wp4256]
>>>
>>> The fix is trivial and obvious, and line 912 must change its JNI
>>> function from "CallObjectMethod"
>>> to "CallBooleanMethod."
>>
>> The patch looks sound.  Would you write a log message, per the patch
>> submission guidelines[1]?  You can find our log message format at [2].
>>
>> Thanks!
>>
>> -Hyrum
>>
>> [1] http://subversion.apache.org/docs/community-guide/general.html#patches
>> [2]
>> http://subversion.apache.org/docs/community-guide/conventions.html#log-messages
>>
>
Received on 2010-05-19 14:57:57 CEST

This is an archived mail posted to the Subversion Dev mailing list.