Re: [Junit-devel] "No runnable methods" exception in org.junit.internal.runners.TestClass

[Curt A. <ca...@ap...>]

On Mar 15, 2006, at 9:30 AM, David Saff wrote:

> Curt Arnold wrote:
>> The Gump (http://gump.apache.org) nightly builds for log4j have  
>> been failing for the last few nights due to a suspected change in  
>> the junit source code.
>
> Sorry about that.  Let's get it fixed.
>
>> I attempted to check out the current CVS HEAD, but the system was  
>> unresponsive.
>
> You're not the first to be frustrated by sourceforge's glacial  
> anonymous cvs access over the last few months.  I'm looking into  
> whether we can get any kind of better guarantees from SourceForge.   
> My apologies.
>
>> The lack of activity on the mailing list archive is also  
>> disconcerting.
>
> Is there a question that has gone unanswered?
>

No, just that it looked like there may also be technical issues with  
the mailing list.  I reviewed the mailing list archive to see if  
there had been recent commits to the CVS that might be related to the  
change of behavior.  There weren't any, but their did not seem to be  
any messages at all in the archives since mid-February which seemed  
suspicious.  Has there been CVS activity in the last month,  
particularly in the last few days or in runners.TestClass?  If not,  
then I might be misattributing the problem.


>> I've updated the Gump metadata so that log4j builds against junit  
>> 3.8.1 not the CVS HEAD and should know overnight if that avoids  
>> the symptoms for now.
>
> You may want to try 4.0 (tagged r40) or 3.8.2 (tagged r382), which  
> are the supported release versions.
>

Gump has project descriptors for junit3 and junit.  The junit3  
project uses a release version of junit.  I thought it was 3.8.1, but  
maybe it has been updated to 3.8.2.  junit builds from the CVS HEAD.   
There is not a project descriptor for a junit 4 release.  I bungled  
the update of the log4j Gump descriptor to use junit3 last night, but  
have hopefully fixed it now.  There is a gump build in process and  
should know in a few hours if reverting to junit 3 suppresses the  
problem.


>> The Gump build fails with:
>>
>>     [junit] Testcase: org.apache.log4j.CoreTestSuite took 0.004 sec
>>     [junit]     Caused an ERROR
>>     [junit] No runnable methods
>>     [junit] java.lang.Exception: No runnable methods
>>
>> CoreTestSuite.java looks like:
>
> [snip]
>
>> What I assume is happening is that one of the classes in the calls  
>> to addTestSuite does not contain any test methods or the  
>> inadvertent (and just recognized) duplication of LoggingEventTest  
>> is resulting in an exception that did not occur with junit prior  
>> to a few days ago.  In either case, the "No runnable methods"  
>> message is of very little help in analyzing the problem.  It would  
>> be helpful at least if the "No runnable methods" message would  
>> include the name of the class that was being examined for runnable  
>> methods.
>
> Yuck.  Can you give me more information about the ant task that is  
> running the tests?  I'd love to know why ant has decided to filter  
> out all but the least useful information provided by JUnit.

Should be running the stock junit task in the CVS HEAD of Ant.  The  
project file is http://svn.apache.org/repos/asf/logging/log4j/trunk/ 
tests/build.xml.  The failure occurs during the "Core" target.

>
>> Unfortunately log4j is a bear to set up to build since it involves  
>> a large number of Sun and third-party dependencies.  I haven't  
>> been able to reproduce the problem on my local machines using  
>> log4j jars from Gump.
>
> So when you run it locally, the tests appear to pass?

Yes, either when I use Ant 1.6.5 and Junit 3.8.1 or a junit.jar from  
Gump.  I have not attempted to use a SVN HEAD version of Ant, though  
it could be the problem.  However, the test "No runnable methods"  
does appear in the TestClass.class file.