I am looking at the application itest failures now and try to fix it.

Many thanks and kindest regards,
Emily
===========================
Emily Jiang





From:   Joe Bohn <[email protected]>
To:     [email protected]
Date:   16/09/2010 13:03
Subject:        Re: building Apache Aries trunk from the top level pom




That's my understanding as well regarding -fae.  It isn't that maven 
does anything different as it builds - it's just that using -fae or -fn 
causes the build to continue even when a failure is encountered.  In the 
scenario I spelled out below (without using -fae or -fn) the failure 
causes the build to terminate immediately.  You only get a complete 
image of all modules when everything succeeds in a single run (which 
practically never happens for me but seems to regularly happen on Hudson).

Joe

On 9/16/10 12:51 AM, Alasdair Nottingham wrote:
> As I understand it using -fae just says keep going if a failure occurs, 
it doesn't change the way maven works.
>
> I ran with MFN clean install 4-5 times and other than the application 
tests everything worked. I'll try with mvn install tomorrow and let you 
know though.
>
> Alasdair Nottingham
>
> On 15 Sep 2010, at 20:04, Joe Bohn<[email protected]>  wrote:
>
>>
>> Thanks for the input.
>>
>> I think building with -fn or -fae is somewhat of a different scenario 
and masks the problem.  If -fn or -fae are necessary then we should update 
the wiki to indicate this.
>>
>> In my case I was trying to run with just mvn install (per our 
directions on the wiki and what I think Hudson is doing).  When a failure 
is encountered I run mvn install again on the top level pom (the 
definition of insanity - doing the same thing and expecting different 
results :-) ).  This process gets you into a potentially never ending 
cycle of failures as tests that passed on the earlier attempts fail on 
subsequent attempts.
>>
>> For example, here are some real life results:
>> - mvn clean from root
>> - remove aries artifacts from maven local repo to ensure a clean 
starting point (rm -rf ./m2/repository/org/apache/aries )
>> - mvn install from root
>> - failures in application tests - IsolatedRuntimeTest, UpdateAppTest, 
OBRResolverTest, and OBRResolverAdvancedTest
>> - mvn install from root
>> - failures in application tests - IsolatedRuntimeTest, OBRResolverTest 
and OBRResolverAdvancedTest
>> - mvn install from root
>> - failures in blueprint tests - BlueprintContainer2Test and 
QuiesceBlueprintTest (NOTE: didn't even get to application this time)
>> - mvn install from root
>> - failure in blueprint tests - QuiesceBlueprintTest
>> - mvn install from root
>> - failure in transaction tests - InvalidTranAttributeTest
>> - mvn install from root
>> - failure in blueprint tests - BlueprintAnnotationTest
>> - mvn install from root
>> - failure in blueprint test - QuiesceBlueprintTest
>> - on and on it goes ... where it stops ... nobody knows :-)
>>
>> I honestly don't know how the Hudson builds are working.  It seems that 
the best anyone on this thread has reported thus far is 50% success (and 
I'm not sure if that was for a successful top level build as I was 
attempting and Hudson appears to accomplish most of the time or if it was 
rebuilding just a failing module individually).
>>
>> Joe
>>
>>
>> On 9/15/10 7:25 PM, Alasdair Nottingham wrote:
>>> So I've just run the build many times with the following:
>>>
>>> export MAVEN_OPTS=-Xmx512m
>>> mvn -fae clean install
>>>
>>> and the only failures I get are in the Aries Application integration
>>> tests component, I haven't looked at details. I'm running on a mac
>>> with java 6.
>>>
>>> Alasdair
>>>
>>> On 15 September 2010 09:42, Emily Jiang<[email protected]>   wrote:
>>>> For me, two areas are constantly failing: blueprint itests:
>>>> quiesceBlueprintTests particularly failed very often followed by some 
of
>>>> application itests.
>>>>
>>>> Many thanks and kindest regards,
>>>> Emily
>>>> ===========================
>>>> Emily Jiang
>>>> WebSphere ESB Foundation Technologies
>>>>
>>>> MP 211, DE3A25, Winchester, Hampshire, England, SO21 2JN
>>>> Phone:  +44 (0)1962 816278  Internal: 246278
>>>>
>>>> Email: [email protected]
>>>> Lotus Notes: Emily Jiang/UK/i...@ibmgb
>>>>
>>>>
>>>>
>>>>
>>>> From:   Mark Nuttall<[email protected]>
>>>> To:     [email protected]
>>>> Date:   15/09/2010 17:20
>>>> Subject:        Re: building Apache Aries trunk from the top level 
pom
>>>>
>>>>
>>>>
>>>> I too find the application itests particularly flaky on a local mvn
>>>> command
>>>> line. I spent most of today trying to debug the sometimes-failing 
tests: I
>>>> couldn't get any to fail under a debugger, and couldn't work out from 
the
>>>> trace why any had failed otherwise :(
>>>>
>>>> Emily, Chris and I are going to be making changes to the application
>>>> provisioning and runtime areas for a while yet. I'm sorry if we've
>>>> introduced further problems in this area. I do think it's timing 
related
>>>> based on today's investigations.
>>>>
>>>> Mark
>>>>
>>>> On 15 September 2010 17:03, Valentin Mahrwald
>>>> <[email protected]>wrote:
>>>>
>>>>> I just checked and had a 50% success rate :)
>>>>>
>>>>> I have found some of the new tests in the application itest project 
are
>>>>> flaky (maybe the timeouts are not big enough). But the rest seems to
>>>> work
>>>>> for me.
>>>>>
>>>>> Valentin
>>>>>
>>>>>
>>>>> On 15 Sep 2010, at 10:28, Joe Bohn wrote:
>>>>>
>>>>>>
>>>>>> I seem to be having lots of problems building Apache Aries trunk 
from
>>>> the
>>>>> top level pom because of test errors.  And, the more tests we add 
the
>>>> worse
>>>>> the process becomes.  For me it is virtually impossible to build. 
Once
>>>> in a
>>>>> while I'll get lucky and things will actually work. However, most of 
the
>>>>> time it seems there are test failures somewhere along the way.  The
>>>> failure
>>>>> is often a timeout waiting for a service. However, there are a large
>>>> number
>>>>> of other (strange) failures such as InvocationTargetExceptions, 
invalid
>>>>> state, NPEs, etc...  that are becoming more common.
>>>>>>
>>>>>> When attempting to run a build from the top level a test that 
passes
>>>> on
>>>>> one attempt will fail on the next and the one that failed on the 
last
>>>> run
>>>>> will pass on the next (if the build even gets that far).  All in 
all, it
>>>> is
>>>>> pretty much impossible to build from the top level.
>>>>>>
>>>>>> The only success that I have in building all of Apache Aries is to
>>>> build
>>>>> each module individually in the order specified in the top level pom
>>>> (which
>>>>> I think is now correct).  As I hit failures I rebuild just that 
module
>>>> until
>>>>> successful and then I move on to the next module.
>>>>>>
>>>>>> So this raises 2 questions:
>>>>>> 1) Am I the only one seeing these types of problems?  If it is just 
me
>>>>> then I guess I just need to figure out what is wrong with my
>>>> environment.
>>>>>>
>>>>>> 2) If it is more wide spread then it seems to me that we might have
>>>>> issues that we need to address.  Certainly we are dealing with a 
dynamic
>>>>> system with loose coupling and there are very likely timing 
scenarios
>>>> that
>>>>> will arise occasionally.  However, the frequency and variety of 
failures
>>>> I'm
>>>>> seeing makes me wonder if we have larger timing or synchronization
>>>> issues
>>>>> that have not yet been addressed.  Do you agree?  If so, then we 
need to
>>>>> come up with some way to isolate and resolve these issues.
>>>>>>
>>>>>> Joe
>>>>>
>>>>>
>>>>
>>>>
>>>>
>>>>
>>>>
>>>>
>>>>
>>>> Unless stated otherwise above:
>>>> IBM United Kingdom Limited - Registered in England and Wales with 
number
>>>> 741598.
>>>> Registered office: PO Box 41, North Harbour, Portsmouth, Hampshire 
PO6 3AU
>>>>
>>>>
>>>>
>>>>
>>>>
>>>>
>>>
>>>
>>>
>>
>>
>> --
>> Joe
>


-- 
Joe







Unless stated otherwise above:
IBM United Kingdom Limited - Registered in England and Wales with number 
741598. 
Registered office: PO Box 41, North Harbour, Portsmouth, Hampshire PO6 3AU





Reply via email to