I think we should take it case-by-case. Guava universal shading is already
quite a mess - build super slow, copies of Guava in jars that don't even
depend on Guava, things shaded that break API surfaces. In this case
there's no mystery what needs to happen, is there?

Kenn

On Fri, Feb 9, 2018 at 11:47 AM, Lukasz Cwik <lc...@google.com> wrote:

> Shading netty everywhere would prevent conflicts with user supplied
> versions of netty.
>
> On Fri, Feb 9, 2018 at 1:03 AM, Ismaël Mejía (JIRA) <j...@apache.org>
> wrote:
>
>>
>>     [ https://issues.apache.org/jira/browse/BEAM-3519?page=com.
>> atlassian.jira.plugin.system.issuetabpanels:comment-tabpane
>> l&focusedCommentId=16358132#comment-16358132 ]
>>
>> Ismaël Mejía commented on BEAM-3519:
>> ------------------------------------
>>
>> The explicit issue probably existed before (GCPIO leaking netty), but it
>> was not detected because the previous version of the Spark runner (based on
>> Spark 1.x) used a version of netty that didn't complain.
>> In any case shading correctly GCPIO would be great to avoid this issue in
>> the future in other runners too.
>>
>> > GCP IO exposes netty on its API surface, causing conflicts with runners
>> > -----------------------------------------------------------------------
>> >
>> >                 Key: BEAM-3519
>> >                 URL: https://issues.apache.org/jira/browse/BEAM-3519
>> >             Project: Beam
>> >          Issue Type: Bug
>> >          Components: sdk-java-gcp
>> >            Reporter: Ismaël Mejía
>> >            Assignee: Chamikara Jayalath
>> >            Priority: Blocker
>> >             Fix For: 2.3.0
>> >
>> >
>> > Google Cloud Platform IOs module leaks netty this causes conflicts in
>> particular with execution systems that use conflicting versions of such
>> modules.
>> >  For the case there is a dependency conflict with the Spark Runner
>> version of netty, see: BEAM-3492
>>
>>
>>
>> --
>> This message was sent by Atlassian JIRA
>> (v7.6.3#76005)
>>
>
>

Reply via email to