maven-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Stephen Connolly <stephen.alan.conno...@gmail.com>
Subject Re: [VOTE] Release Apache Maven 3.5.1
Date Fri, 15 Sep 2017 09:54:54 GMT
On 14 September 2017 at 23:55, Michael Osipov <michaelo@apache.org> wrote:

> Am 2017-09-15 um 00:50 schrieb Petr Široký:
>
>> I was able to easily fix our plugin by e.g. replacing
>> "Thread.currentThread().getContextClassLoader()" with
>> "this.getClass().getClassLoader()" (in the Mojo class) to get the plugin
>> classloader.
>>
>> I don't know though if the "Thread.currentThread().getCon
>> textClassLoader()"
>> is just misuse on our side or if it's something that more plugins may rely
>> on.
>>
>
> Similar cause in MASSEMBLY: https://issues.apache.org/jira/browse/MNG-6209
>
> I think using TCCL is wrong here.


I agree, I think the TCCL is supposed to be for extension lookup.

If you want the plugin's classloader then that should be
`this.getClass().getClassLoader()` from the Mojo class.

If you want the classloader with all the extensions, that should be TCCL.

What is unclear to me is why we set TCCL to anything other than the project
realm. A pom plugin declaration of <extensions>true</extensions> should not
affect the TCCL that the plugin execution sees, so I wonder if the root bug
is that the fix allows the TCCL to be other than project realm?

Igor?


>
>
> On Thu, Sep 14, 2017 at 2:42 PM Petr Široký <siroky.petr@gmail.com> wrote:
>>
>> Argh, I forgot to link the plugin source:
>>> https://github.com/kiegroup/droolsjbpm-integration/tree/7.3.
>>> 0.Final/kie-maven-plugin
>>>
>>> On Thu, Sep 14, 2017 at 2:41 PM Petr Široký <siroky.petr@gmail.com>
>>> wrote:
>>>
>>> Hello,
>>>>
>>>> I am seeing a (probably) similar issue with our custom plugin.
>>>>
>>>> See the reproducer:
>>>> https://github.com/psiroky/reproducers/tree/mvn351-kie-maven-plugin
>>>> (works
>>>> fine with maven 3.5.0, but fails with NPE with the RC of maven 3.5.1).
>>>>
>>>> I am not yet sure if the plugin is just doing something it's not
>>>> supposed
>>>> to, or if this is a regression in maven itself. I'll will take a deeper
>>>> look.
>>>>
>>>> Petr
>>>>
>>>>
>>>> On Thu, Sep 14, 2017 at 1:53 PM Stephen Connolly <
>>>> stephen.alan.connolly@gmail.com> wrote:
>>>>
>>>> On 14 September 2017 at 04:43, Mark Derricutt <mark@talios.com> wrote:
>>>>>
>>>>> +2 non-binding from Mark!
>>>>>>>
>>>>>>
>>>>>> I was discussing this with a coworker and he made the comment that
if
>>>>>>
>>>>> this
>>>>>
>>>>>> change could break Mojos, maybe it shouldn't be in a point release
-
>>>>>>
>>>>> whats
>>>>>
>>>>>> the policy on changes that may potentially break existing plugins?
>>>>>>
>>>>>>
>>>>> Well we need to assess the issue. Right now I don't even have a
>>>>> description
>>>>> of what went wrong. Any chance you could provide a replication... or
>>>>> mail
>>>>> me directly if you cannot share it publically and I may be able to
>>>>> produce
>>>>> a minimal reproduction from it.
>>>>>
>>>>> If this breaks a mojo that was doing something wrong in the first
>>>>> place,
>>>>> well that will not stop 3.5.1... OTOH if this exposes a bug in the
>>>>> issue
>>>>> "fixed" then I'd likely revert and respin.
>>>>>
>>>>> We really need a reproducer first.
>>>>>
>>>>>
>>>>>
>>>>>> --
>>>>>> "Great artists are extremely selfish and arrogant things" — Steven
>>>>>>
>>>>> Wilson,
>>>>>
>>>>>> Porcupine Tree
>>>>>>
>>>>>> On Thu, Sep 14, 2017 at 10:29 AM, Mark Derricutt <mark@talios.com>
>>>>>>
>>>>> wrote:
>>>>>
>>>>>>
>>>>>> On 14 Sep 2017, at 10:26, Mark Derricutt wrote:
>>>>>>>
>>>>>>> Calling -2 for vote if not too late.
>>>>>>>
>>>>>>> Actually - looking at the commit diff, I see in our code we did
have
>>>>>>> <extensions>true</extensions> for the jasmine-maven-plugin
which we
>>>>>>>
>>>>>> don't
>>>>>
>>>>>> actually need. Removing that from the mojo definition and running
my
>>>>>>>
>>>>>> build
>>>>>>
>>>>>>> with the staged 3.5.1 release and everything builds fine.
>>>>>>>
>>>>>>> +2 non-binding from Mark!
>>>>>>>
>>>>>>> Mark
>>>>>>> ------------------------------
>>>>>>>
>>>>>>> "The ease with which a change can be implemented has no relevance
at
>>>>>>>
>>>>>> all
>>>>>
>>>>>> to whether it is the right change for the (Java) Platform for all
>>>>>>>
>>>>>> time."
>>>>>
>>>>>> —
>>>>>>
>>>>>>> Mark Reinhold.
>>>>>>>
>>>>>>> Mark Derricutt
>>>>>>> http://www.theoryinpractice.net
>>>>>>> http://www.chaliceofblood.net
>>>>>>> http://plus.google.com/+MarkDerricutt
>>>>>>> http://twitter.com/talios
>>>>>>> http://facebook.com/mderricutt
>>>>>>>
>>>>>>>
>>>>>>
>>>>>
>>>>
>>
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: dev-unsubscribe@maven.apache.org
> For additional commands, e-mail: dev-help@maven.apache.org
>
>

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message