uima-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Marshall Schor <...@schor.com>
Subject Re: something strange with Jenkins builds, test case failures
Date Wed, 05 Nov 2014 19:21:06 GMT
Running locally on my laptop (Windows) I observe:

1) mvn package -Pfindbugs  - runs with findbugs, both 3.0.0 (newer version) and
2.5.4 (version being used), but findbugs doesn't print any indication like it
does on Jenkins that it's recursively rebuilding and re-running tests.

I saw that my mvn version is 3.0.5, while the Jenkins was running 3.0.3.  So I
downgraded my laptop mvn version to 3.0.3 - the mvn build still runs fine.

  - this means I can't reproduce this on my laptop.  It's likely that there's
some interaction with Jenkins and Maven which is causing this.

2) Just incidentally, I noticed that findbugs fails when running with Oracle
Java 8 (1.8.0_25-b18)

message: [java]   Unable to get XClass for java/lang/StringBuffer
[java]     java.lang.ArrayIndexOutOfBoundsException: 26721
[java]       At org.objectweb.asm.ClassReader.readClass(Unknown Source) etc.

it works OK with IBM Java 8 (build pwa6480ea-20130422_01)

----------
I've restored the build to Ubuntu || Windows, and changed the maven level to
"latest" (was 3.0.3).

I'll probably close the issues as Not a problem, since the errors appear to be
specific for findbugs running in the Jenkins environment.

-Marshall

On 11/5/2014 10:57 AM, Marshall Schor wrote:
> Here's what's happening; it happens on both Windows1 and Ubuntu recent builds.
>
>
> The uimaj-core (where the "failing" test are) is built normally, no errors
>
>    - the Java is OK (
>        Oracle 1.7.0 Java HotSpot(TM) 64-Bit Server VM 21.0-b17 on Windows 1 and
>        Oracle 1.7.0_25 Java HotSpot(TM) Server VM 23.25-b01 on Ubuntu)
>
> However, the build on Jenkins includes a flag that causes Jenkins to run the
> "Findbugs" maven plugin.
>
> Findbugs maven plugin running seems unusual.  It causes a recursive build of the
> module, which not only recompiles everything, but re-runs the tests as well.  It
> is only this second running of the tests that fail.  It's likely that findbugs
> configuration is somehow specifying an older version of Xalan that ends up not
> supporting XML 1.1.
>
> Investigating further to see if we can configure Findbugs to not re-do the
> compilation and rerunning, and/or to fix its version of Xalan.
>
> -Marshall
>
>
> On 11/5/2014 10:38 AM, Marshall Schor wrote:
>> So, try # 1 was assigned "Windows2" - a different Windows machine than used
>> before (in build 586).  That build
>> didn't even get started - while Jenkins was parsing the maven poms, it threw a
>> fatal "out of permgen space". 
>> It sounds like the Java level installed on that machine has some configuration
>> issues.
>>
>> I restarted it, and it is assigned now to Windows1...
>>
>> -Marshall
>> On 11/5/2014 10:25 AM, Marshall Schor wrote:
>>> I added some debug output to record the JVM name and the specified name of the
>>> TransformerFactory (if any).
>>>
>>> I reran on Jenkins, and Jenkins picked the Ubuntu slave (failure was on
>>> Windows1) and I watched the console - no error reported on the xml tests.
>>>
>>> I've temporarily restricted the build for UIMA-SDK to "Windows" to see if I can
>>> reproduce this failure.
>>>
>>> -Marshall
>>>
>>> On 11/4/2014 4:43 PM, Marshall Schor wrote:
>>>> Hi,
>>>>
>>>> I was watching the console log for UIMA-SDK #586, and noticed it said along
the
>>>> way some failure indications, for instance:
>>>>
>>>> testCAStoString(org.apache.uima.util.CasToInlineXmlTest)  Time elapsed: 0.027
>>>> sec  <<< FAILURE!
>>>>
>>>> So, I was quite surprised when the build finished and sent email to this
list
>>>> saying everything was successful.
>>>>
>>>> Going to Jenkins test report for build # 586 here:
>>>> https://builds.apache.org/job/UIMA-SDK/org.apache.uima$uimaj-core/586/testReport/
says
>>>> there are no errors.
>>>>
>>>> But looking at the console output
>>>> https://builds.apache.org/view/All/job/UIMA-SDK/586/console definitely shows
>>>> errors. 
>>>>
>>>> (The errors seem to be due to different XML formatters; one writing <xxx
 ....
>>>> /> and the other <xxx .... ></xxx>. Richard pointed me to
a utility to get
>>>> around this, and I can add that, so these won't fail.)
>>>>
>>>> But more importantly, does anyone have any idea why one part of Jenkins (the
>>>> console log) is reporting failures, and the other part (test summary) is
saying
>>>> there are no failures?
>>>>
>>>> -M
>>>>
>>>>
>>
>
>


Mime
View raw message