git.net

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Test failures in master


I've created DAFFODIL-1945 [1] to track this. I wouldn't recommend
trying to fix it just yet, but if you want to move those tests to
src/test/scala-debug and verify that all your tests pass after that
change, that would be relatively straight forward first contribution.

- Steve

[1] https://issues.apache.org/jira/browse/DAFFODIL-1945

On 05/24/2018 01:27 PM, Steve Lawrence wrote:
> It's up to you, but I can imagine this is going to be a painful place to
> debug. We've had plenty of issues with the ICU library in the past and
> it's never fun.
> 
> I'd recommend starting with one of the issues previously mentioned, at
> least to get your feet wet with Daffoil development.
> 
> I'll create a bug to track the issues you found and we'll move them to
> scala-debug until it's sorted out. If you can think of any thing that
> might be helpful in reproducing this issue, feel free to email or add it
> to the bug.
> 
> - Steve
> 
> 
> On 05/24/2018 12:12 PM, Russ Williams wrote:
>> Cheers! Yeah, I’m not familiar with SBT, so that’ll come in handy.
>>
>> If it were a thread issue, I’d expect the results to be more random - the failures have been 100% consistent, in the same tests, with the same values returned. Could just be that I’m lucky.
>>
>> I’ll also install Eclipse - have used it a lot, but not on my home machine - and see if I can debug into the test to see why the results are coming out the way they are.
>>
>> Given that this is (currently) only affecting me, shall I ignore these specific tests and go ahead with the branch/bug fixes as discussed yesterday?
>>
>>> On 24 May 2018, at 14:06, Steve Lawrence <slawrence@xxxxxxxxxx> wrote:
>>>
>>> That's a good thought. If you're unfamiliar with sbt, to run a single
>>> test you can do something like this:
>>>
>>> $ sbt "testOnly org.apache.daffodil.IBMTestsThatPass --
>>> --tests=test_simple_type_properties_text_calendar_13_01"
>>>
>>> - Steve
>>>
>>> On 05/24/2018 08:48 AM, Mike Beckerle wrote:
>>>> Russ,
>>>>
>>>> The main thing I hate about the java time libraries (old school Date, Time, DateTime) is that they are stateful, when run en-masse, our tests are multi-threaded.
>>>>
>>>> Have you tried running these tests one at a time in isolation? I am wondering if someplace we are sharing state accidently?
>>>>
>>>> ...mikeb
>>>>
>>>> ________________________________
>>>> From: Russ Williams <russ@xxxxxxxxxxxxxxx>
>>>> Sent: Thursday, May 24, 2018 4:47:52 AM
>>>> To: dev@xxxxxxxxxxxxxxxxxxx
>>>> Subject: Test failures in master
>>>>
>>>> Hi!
>>>>
>>>> Testing the build process with a clone of the current master branch, I’m getting five test failures related to calendar handling for day-of-week and week-of-year.
>>>>
>>>> One in daffodil-test-ibm1/test:
>>>>        [error] Test org.apache.daffodil.IBMTestsThatPass.test_simple_type_properties_text_calendar_13_01 failed: java.lang.Exception:
>>>>        [error] Comparison failed.
>>>>        [error] Expected
>>>>        [error]           <myDateTime>2010-12-27T04:05:06.000000+00:00</myDateTime>
>>>>        [error] Actual
>>>>        [error]           <myDateTime>2010-12-20T04:05:06.000000+00:00</myDateTime>
>>>>
>>>>
>>>> Four in daffodil-test/test:
>>>>        [error] Test org.apache.daffodil.section05.simple_types.TestSimpleTypes.test_dateCalendarDaysInFirstWeek3 failed: java.lang.Exception:
>>>>        [error] Comparison failed.
>>>>        [error] Expected
>>>>        [error]           <date17>2012-01-01+00:00</date17>
>>>>        [error] Actual
>>>>        [error]           <date17>2012-12-23+00:00</date17>
>>>>
>>>>        [error] Test org.apache.daffodil.section05.simple_types.TestSimpleTypes.test_dateCalendarDaysInFirstWeek5 failed: java.lang.Exception:
>>>>        [error] Comparison failed.
>>>>        [error] Expected
>>>>        [error]           <date20>2013-02-24+00:00</date20>
>>>>        [error] Actual
>>>>        [error]           <date20>2013-02-10+00:00</date20>
>>>>
>>>>        [error] Test org.apache.daffodil.section05.simple_types.TestSimpleTypes.test_dateCalendarFirstDayOfWeek03 failed: java.lang.Exception:
>>>>        [error] Comparison failed.
>>>>        [error] Expected
>>>>        [error]           <date06>2013-02-03+00:00</date06>
>>>>        [error] Actual
>>>>        [error]           <date06>2013-02-02+00:00</date06>
>>>>
>>>>        [error] Test org.apache.daffodil.section05.simple_types.TestSimpleTypes.test_dateCalendarFirstDayOfWeek04 failed: java.lang.Exception:
>>>>        [error] Comparison failed.
>>>>        [error] Expected
>>>>        [error]           <date06>2013-02-04+00:00</date06>
>>>>        [error] Actual
>>>>        [error]           <date06>2013-02-03+00:00</date06>
>>>>
>>>>
>>>> The dateCalendarDaysInFirstWeek3 failure is particularly bad since it’s looking for “week 1 of 2012”, which should start on 2012-01-01, but it’s getting a date in *December* 2012.
>>>>
>>>>
>>>> Every other test passes cleanly.
>>>> I’ve got the same result on my Mac (macOS Sierra/10.12.6, Oracle Java 1.8.0_51-b16) and a Linux box (Ubuntu 16.04 LTS, Oracle Java 1.8.0_111-b14).
>>>> Setting -Duser.timezone in JAVA_OPTS doesn’t make any difference.
>>>> I haven’t been able to test with OpenJDK yet.
>>>>
>>>> Cheers,
>>>> —
>>>> Russ
>>>>
>>>
>>
>