I’m going to punt on this. For now, let’s just remove the unit tests. We’ll 
have the history in git and can resurrect the code if the licensing situation 
changes.

Although I was able to get the tests to pass with an in-memory Cassandra 
instance, I was not able to get them to pass reliably and consistently. The 
tests were sensitive to timing and the performance of the build machine. We’ve 
made some great progress addressing some of the causes of spurious test 
failures, and I’d hate to take a step backward.

I contacted the author of cassandra-unit regarding changing the license, and he 
seems open to discussing a change in licensing [1], so there’s a chance the 
situation may change.

[1] https://github.com/jsevellec/cassandra-unit/issues/163 
<https://github.com/jsevellec/cassandra-unit/issues/163>

-Taylor


> On Jan 8, 2016, at 12:23 PM, Bobby Evans <[email protected]> wrote:
> 
> Great to hear it.  I will wait for this to go in before I create a 1.0 
> branch.  Not reason to not wait.  So now I expect to create the branch 
> sometime Monday.
>  - Bobby
> 
>    On Thursday, January 7, 2016 2:47 PM, P. Taylor Goetz <[email protected]> 
> wrote:
> 
> 
> Okay, I just got all the storm-cassandra tests to pass without cassandra-unit 
> by writing a quick-and-dirty embedded C* instance.
> 
> I’ve got some cleanup and refactoring to do, but it looks like a viable 
> alternative. I should have a pull request up soon.
> 
> -Taylor
> 
>> On Jan 7, 2016, at 2:55 PM, Bobby Evans <[email protected]> wrote:
>> 
>> That seems reasonable to me.  Thanks for looking into this Taylor.
>>   - Bobby
>> 
>>     On Wednesday, January 6, 2016 4:28 PM, P. Taylor Goetz 
>> <[email protected]> wrote:
>> 
>> 
>> More information:
>> 
>> The fact that the code is in our repo is not a problem, it just means that 
>> we we can’t release until that is rectified by removing the LGPL dependency. 
>> When Storm first entered incubation we had a dependency on 0mq (LGPL), which 
>> didn’t stop the code from being imported into the Apache repo, but it had to 
>> be removed before we could officially release.
>> 
>> Looking at the unit tests in question, it doesn’t seem like it would be hard 
>> to migrate away from cassandra-unit. I’ve used cassandra-unit in the past (a 
>> while ago), and eventually migrated away from it by spinning up an in-memory 
>> C* instance and populating it with the necessary data from within the unit 
>> tests, which is what cassandra-unit seems to be used for here.
>> 
>> Florian — would you be able to help migrate off of cassandra-unit, possibly 
>> considering the approach I mentioned? If not I may be able to find some time 
>> to do it.
>> 
>> If for some reason we can’t migrate by the time we’re ready to release, we 
>> can just delete the tests/dependency. But I’d at least try to 
>> preserve/migrate them since I believe they have value.
>> 
>> -Taylor
>> 
>> 
>>> On Jan 6, 2016, at 4:13 PM, P. Taylor Goetz <[email protected]> wrote:
>>> 
>>> Let me do some quick research before you rip anything out.
>>> 
>>> -Taylor
>>> 
>>>> On Jan 6, 2016, at 4:10 PM, Bobby Evans <[email protected]> 
>>>> wrote:
>>>> 
>>>> Yes lets just remove them for now, and then file a follow up JIRA to add 
>>>> back in tests.
>>>> - Bobby
>>>> 
>>>>   On Wednesday, January 6, 2016 3:01 PM, Jungtaek Lim <[email protected]> 
>>>> wrote:
>>>> 
>>>> 
>>>> Sorry missed link, https://issues.apache.org/jira/browse/STORM-1445
>>>> 
>>>> 2016년 1월 7일 (목) 오전 5:59, Jungtaek Lim <[email protected]>님이 작성:
>>>> 
>>>>> Filed STORM-1445.
>>>>> 
>>>>> Seems like unit tests are completely relying on cassandra-unit.
>>>>> I don't have experience with Cassandra, so I couldn't convert current
>>>>> tests to not use cassandra-unit.
>>>>> 
>>>>> If we think we're fine to remove whole unit tests for storm-cassandra,
>>>>> I'll remove it and submit pull request right now.
>>>>> If we still need unit tests for storm-cassandra, I'd love to let sponsors
>>>>> of storm-cassandra module takes care of it.
>>>>> (Maybe we can file a new issue which handles new unit tests.)
>>>>> 
>>>>> Best,
>>>>> Jungtaek Lim (HeartSaVioR)
>>>>> 
>>>>> 
>>>>> 
>>>>> 2016년 1월 7일 (목) 오전 5:31, Bobby Evans <[email protected]>님이 작성:
>>>>> 
>>>>>> Yes pull it out for now, and we may have to talk to someone in legal at
>>>>>> apache if there is something else we need to do.  We have not done a
>>>>>> release with storm-cassandra yet, so we are probably safe.  Because of 
>>>>>> that
>>>>>> please file a JIRA and put up a pull request like normal.
>>>>>>   - Bobby
>>>>>> 
>>>>>>     On Wednesday, January 6, 2016 7:35 AM, Jungtaek Lim <
>>>>>> [email protected]> wrote:
>>>>>> 
>>>>>> 
>>>>>>   Hi devs,
>>>>>> 
>>>>>> Digging into the test failures on storm-cassandra, I saw license of
>>>>>> cassandra-unit is LGPL v3 by chance.
>>>>>> 
>>>>>> https://github.com/jsevellec/cassandra-unit
>>>>>> 
>>>>>> From http://www.apache.org/legal/resolved.html, the page describes that
>>>>>> 
>>>>>> LGPL-licensed works must not be included in Apache Products
>>>>>> 
>>>>>> 
>>>>>> but I don't know much details on license so clarification would be much
>>>>>> appreciated.
>>>>>> 
>>>>>> Should we get rid of cassandra-unit? I sought the alternatives, but
>>>>>> nothing
>>>>>> found.
>>>>>> 
>>>>>> Best,
>>>>>> Jungtaek Lim (HeartSaVioR)
>>>>>> 
>>>>>> --
>>>>>> Name : Jungtaek Lim
>>>>>> Blog : http://medium.com/@heartsavior
>>>>>> Twitter : http://twitter.com/heartsavior
>>>>>> LinkedIn : http://www.linkedin.com/in/heartsavior
>>>>>> 
>>>>>> 
>>>>>> 
>>>>> 
>>>>> 
>>>> 
>>> 
>> 
>> 
> 
> 

Attachment: signature.asc
Description: Message signed with OpenPGP using GPGMail

Reply via email to