+1 to go ahead with a source release

Sent from my iPhone

> On Aug 19, 2016, at 11:52 PM, Ellison Anne Williams 
> <[email protected]> wrote:
> 
> After spending a bit of time taking a look at the submodule breakout, I
> would rather wait and break the project out into submodules in a more
> deliberate manner (not as a first pass just for the binary artifacts).
> 
> As submodules seem to be required to get the L&N files for binary
> distributions up to par, I propose that we proceed with a source only
> release now. We will slate the submodules for our next release (or so).
> 
> We've certainly learned tons about the release process over the last week
> and I think that it would be good to go ahead and exercise the full process
> to completion.
> 
> Unless anyone objects, I will try to get the source release artifacts up
> for voting over the weekend.
> 
> Thanks!
> 
> On Fri, Aug 19, 2016 at 3:58 PM, Ellison Anne Williams <
> [email protected]> wrote:
> 
>> I may give breaking the project into basic submodules a quick shot over
>> the weekend -- not full submodules, just an assembly and core to get past
>> the release and then we can design from there. If it proves too time
>> consuming, I think that (unfortunately) we should go with a source-only
>> release until we tackle submodules.
>> 
>> On Fri, Aug 19, 2016 at 12:41 PM, Suneel Marthi <[email protected]>
>> wrote:
>> 
>>> I was trying to exclude the logs/ from being packaged into the sources jar
>>> - and seems like the clean way to do it is via assembly.
>>> I don't think its very productive to throw in a hacky fix now and having
>>> to
>>> revert it again to do it the right way.
>>> 
>>> I suggest that we go ahead with this release and pun the highlighted
>>> issues
>>> in the next sprint following this release.
>>> Suneel
>>> 
>>> On Fri, Aug 19, 2016 at 11:57 AM, Ellison Anne Williams <
>>> [email protected]> wrote:
>>> 
>>>> Exactly - which requires a submodule refactor, hence holding off...
>>>> 
>>>> Any other way to satisfies the requirements that Tim mentioned in his
>>> last
>>>> email?
>>>> 
>>>> On Fri, Aug 19, 2016 at 11:55 AM, Suneel Marthi <
>>> [email protected]>
>>>> wrote:
>>>> 
>>>>> I have been looking at this, seems like most of this is best handled
>>> by
>>>>> maven-assembly-plugin - packaging licenses into bin.jar and src.jar in
>>>> the
>>>>> appropriate way; excluding logs/ from source jar etc..
>>>>> 
>>>>> I guess we need to mark these as JIRA issues and pun them for the
>>> next
>>>>> release.
>>>>> 
>>>>> 
>>>>> 
>>>>> On Fri, Aug 19, 2016 at 11:48 AM, Ellison Anne Williams <
>>>>> [email protected]> wrote:
>>>>> 
>>>>>> Understood.
>>>>>> 
>>>>>> Does anyone know how to exclude / stop maven from generating the
>>>> license
>>>>>> files and placing them in META-INF? I don't know off of the top of
>>> my
>>>>> head
>>>>>> and would appreciate not having to spend lots of time digging if
>>>> someone
>>>>>> already knows how to do it. Thanks!
>>>>>> 
>>>>>> On Fri, Aug 19, 2016 at 11:26 AM, Tim Ellison <
>>> [email protected]>
>>>>>> wrote:
>>>>>> 
>>>>>>> The fact that "apache-pirk-0.1.0-exe.jar" contains
>>>>>>>        /LICENSE.txt
>>>>>>> 
>>>>>>> which is a BSD license text, and the actual license for this JAR
>>> is
>>>>>>>        /META-INF/bin-license-notice/LICENSE-bin
>>>>>>> 
>>>>>>> is a blocker for me.
>>>>>>> 
>>>>>>> It is not acceptable to say that the JAR does contain the correct
>>>>>>> license somewhere, with some name.
>>>>>>> 
>>>>>>> 
>>>>>>> How about excluding all the other licenses and notices in the exe
>>> jar
>>>>>>> *except* /META-INF/bin-license-notice/**
>>>>>>> 
>>>>>>> Regards,
>>>>>>> Tim
>>>>>>> 
>>>>>>>> On 19/08/16 16:07, Ellison Anne Williams wrote:
>>>>>>>> Comments inline below.
>>>>>>>> 
>>>>>>>> Additionally, as stated before, this is not the most elegant
>>>> solution
>>>>>> but
>>>>>>>> is seems to satisfy all of the L&N ASF requirements (that I can
>>>>> find).
>>>>>> It
>>>>>>>> seems that a full re-factor with submodules (a la NiFi) will be
>>>>>> necessary
>>>>>>>> to make it cleaner. I would prefer that we make cleaning up the
>>> L&N
>>>>>>>> situation a JIRA issue to go hand in hand with the submodule
>>>> refactor
>>>>>> and
>>>>>>>> proceed with the release. If you have any suggestions on
>>> cleaning
>>>>> that
>>>>>>>> doesn't involve a submodule refactor, I happy to implement them.
>>>>>>>> 
>>>>>>>> 
>>>>>>>> On Fri, Aug 19, 2016 at 10:53 AM, Suneel Marthi <
>>>> [email protected]>
>>>>>>> wrote:
>>>>>>>> 
>>>>>>>>> On Fri, Aug 19, 2016 at 10:13 AM, Tim Ellison <
>>>>> [email protected]>
>>>>>>>>> wrote:
>>>>>>>>> 
>>>>>>>>>>> On 19/08/16 14:52, Ellison Anne Williams wrote:
>>>>>>>>>>> Can you please take a look at PR 65 (PIRK-53) and let me
>>> know if
>>>>> we
>>>>>>> are
>>>>>>>>>>> ready to go with L&N for our first release?
>>>>>>>>>> 
>>>>>>>>>> I feel like I'm being the bad cop :-(
>>>>>>>>>> 
>>>>>>>>>> I'm trying to put myself into the shoes of somebody picking up
>>>> one
>>>>> of
>>>>>>>>>> these artefacts, and figuring out what the terms are under
>>> which
>>>> it
>>>>>> was
>>>>>>>>>> received.  Having L&N files in there that are correct but
>>> mixed
>>>> in
>>>>>>>>>> amongst other incorrect L&N files is just too confusing.  How
>>>> can a
>>>>>>> user
>>>>>>>>>> know which ones apply?  So we have to filter out the rotten
>>> ones,
>>>>> and
>>>>>>>>>> place the correct ones where they would expect to be found.
>>>>>>>>>> 
>>>>>>>>>> Looking into each of the ZIP/JAR files produced from
>>>>>>> ellisonanne/pirk-53
>>>>>>>>>> branch, I see the following:
>>>>>>>>>> 
>>>>>>>>>> apache-pirk-0.0.2-source-release.zip (Main source release)
>>>>>>>>>>        correct
>>>>>>>>>>                /LICENSE
>>>>>>>>>>                /NOTICE
>>>>>>>>>>                /DISCLAIMER
>>>>>>>>>>        confusing (but off topic)
>>>>>>>>>>                /logs - contains old log files
>>>>>>>>> 
>>>>>>>>> These need to be excluded when building the artifact.
>>>>>>>>> 
>>>>>>>> 
>>>>>>>> The logs is probably my fault on a sloppy commit - I will clean
>>> it.
>>>>>>> Thanks
>>>>>>>> for pointing it out.
>>>>>>>> 
>>>>>>>> 
>>>>>>>>> 
>>>>>>>>> 
>>>>>>>>>> apache-pirk-0.1.0.jar (Binary project JAR), and
>>>>>>>>>> apache-pirk-0.1.0-sources.jar (Corresponding source for
>>> binary
>>>>>> project
>>>>>>>>>> JAR)
>>>>>>>>>>        correct
>>>>>>>>>>                /META-INF/LICENSE
>>>>>>>>>>                /META-INF/NOTICE
>>>>>>>>>>        incorrect
>>>>>>>>>>                /META-INF/bin-license-notice/licenses/* -
>>> does
>>>> not
>>>>>>>>> relate
>>>>>>>>>> to the
>>>>>>>>>> content of this JAR.
>>>>>>>>>>                /META-INF/bin-license-notice/LICENSE-bin -
>>> does
>>>>> not
>>>>>>>>>> relate to the
>>>>>>>>>> content of this JAR.
>>>>>>>>>>                /META-INF/bin-license-notice/NOTICE-bin -
>>> does
>>>> not
>>>>>>>>> relate
>>>>>>>>>> to the
>>>>>>>>>> content of this JAR.
>>>>>>>>>>        confusing
>>>>>>>>>>                /META-INF/DISCLAIMER - missing, but probably
>>> ok.
>>>>>>> Ideally
>>>>>>>>>> would
>>>>>>>>>> appear in this JAR too.
>>>>>>>> 
>>>>>>>> Yes, the bin-license-notice directory contain the L&N files that
>>>>>>> correspond
>>>>>>>> to the binary distribution not the source release; however, they
>>>> are
>>>>>>>> clearly labeled, so there shouldn't be any user confusion. This
>>>>> doesn't
>>>>>>>> appear to be contradictory to what I see in NiFi - the binary
>>> L&N
>>>>> files
>>>>>>> are
>>>>>>>> in the nifi-assembly directory which gets packaged into the
>>> source
>>>>>>> release.
>>>>>>>> 
>>>>>>>> 
>>>>>>>> 
>>>>>>>>>> apache-pirk-0.1.0-exe.jar (Combined JAR with all dependencies
>>>>>>> included)
>>>>>>>>>>        incorrect
>>>>>>>>>>                /LICENSE.txt - this is a BSD license.
>>>>>>>>>>                /META-INF/LICENSE - does not refer to
>>>> subcomponents
>>>>>>>>>> license/ directory.
>>>>>>>>>>                /META-INF/NOTICE - does not contain required
>>>>> notices
>>>>>>> from
>>>>>>>>>> subcomponents.
>>>>>>>>>>                /META-INF/LICENSE.txt - contains additional
>>>> license
>>>>>>>>>> statements beyond
>>>>>>>>>> ALv2.
>>>>>>>>>>                /META-INF/NOTICE.txt - does not reference Pirk
>>>>>>>>> incubating,
>>>>>>>>>> contains
>>>>>>>>>> ref to LICENSE.txt
>>>>>>>>>>                /META-INF/license/* - appears not the be the
>>> set
>>>> of
>>>>>>>>>> subcomponent
>>>>>>>>>> licenses (count=10)
>>>>>>>>>>                /META-INF/bin-license-notice/LICENSE-bin -
>>> not
>>>>>> called
>>>>>>>>>> LICENSE.
>>>>>>>>>>                /META-INF/bin-license-notice/NOTICE-bin - not
>>>>> called
>>>>>>>>>> NOTICE.
>>>>>>>>>>        confusing
>>>>>>>>>>                /LICENSE-junit.txt - should be in license/
>>>>> directory.
>>>>>>>>>>                /META-INF/bin-license-notice/licenses/* -
>>> should
>>>>> be
>>>>>> in
>>>>>>>>>> /META-INF/license/*
>>>>>>>>>>                /META-INF/bin-license-notice/LICENSE-bin -
>>>> should
>>>>> be
>>>>>>> in
>>>>>>>>>> /META-INF/LICENSE
>>>>>>>>>>                /META-INF/bin-license-notice/NOTICE-bin -
>>> should
>>>>> be
>>>>>> in
>>>>>>>>>> /META-INF/NOTICE
>>>>>>>> 
>>>>>>>> All of the L&N files that you referenced as incorrect are
>>>>> automatically
>>>>>>>> added into the exe jar -- I'm not sure how to prevent this from
>>>>>> happening
>>>>>>>> off of the top of my head (I'm sure it's possible...), but my
>>>>>>> understanding
>>>>>>>> is that it's not a violation of the policy, it's just not the
>>>>> cleanest
>>>>>>>> approach. I didn't see a hard requirement that the binary L&N
>>> files
>>>>> had
>>>>>>> to
>>>>>>>> be called 'LICENSE' or 'NOTICE' - just that they had to be
>>> present
>>>>> and
>>>>>>>> clearly labeled. Happy to change them to be 'LICENSE' and
>>> 'NOTICE'
>>>> in
>>>>>> the
>>>>>>>> bin-license-notice dir, but I would be more confused as a user.
>>>>>>>> 
>>>>>>>> I also didn't see a hard requirement that the 'licenses'
>>> directory
>>>>>> needed
>>>>>>>> to be directly under META-INF, in fact, I think that it's more
>>>>>> confusing
>>>>>>>> that way as it would show up in the source release.
>>>>>>>> 
>>>>>>>> 
>>>>>>>>>> 
>>>>>>>>>> I'd appreciate other mentors' close review of these built
>>>> artefacts
>>>>>>> too.
>>>>>>>>>> 
>>>>>>>>>> Seems that the main project source release looks good, but the
>>>>>>>>>> convenience JARs still needs some filtering/moving work.
>>>>>>>>>> 
>>>>>>>>>> Anything else you can think of. Not sure how much of this can
>>> be
>>>>>>>>> addressed
>>>>>>>>> in this release. Its best to acknowledge the remaining license
>>>>> issues
>>>>>>> that
>>>>>>>>> need to be addressed and move forward with cutting a release.
>>>>>>>>> 
>>>>>>>>> There just doesn't seem a clean way to get all of this
>>>> straightened
>>>>>> out
>>>>>>> for
>>>>>>>>> the very first release, its possible that the way we are
>>> packaging
>>>>> the
>>>>>>>>> artifacts now may not be the right way and we need to do it
>>>>>> differently,
>>>>>>>>> something that we had not considered or talked about so far.
>>>>>>>>> 
>>>>>>>>> Regards,
>>>>>>>>>> Tim
>>>>>>>>> 
>>>>>>>>> @Ellison how are we packaging the licenses ? I agree that the
>>>> logs/,
>>>>>>>>> bin-license-notice/ shuldn't be showing up.
>> 
>> 

Reply via email to