[ 
https://issues.apache.org/jira/browse/FLINK-386?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14076256#comment-14076256
 ] 

ASF GitHub Bot commented on FLINK-386:
--------------------------------------

Github user StephanEwen commented on the pull request:

    https://github.com/apache/incubator-flink/pull/80#issuecomment-50344798
  
    An enhancement could be to exclude some of the hadoop dependencies for 
Hadoop 1. In Hadoop 1, there is no separation between core, mapred, hdfs, 
client, etc. So we have all the dependencies (JSP, jettison, jetty, hsqldb), 
which are most likely not needed.


> Duplicate dependencies in lib folder
> ------------------------------------
>
>                 Key: FLINK-386
>                 URL: https://issues.apache.org/jira/browse/FLINK-386
>             Project: Flink
>          Issue Type: Bug
>            Reporter: GitHub Import
>            Assignee: Daniel Warneke
>              Labels: github-import
>             Fix For: pre-apache
>
>
> There are some dependencies for which two different versions end up in the 
> lib folder
> ```
> servlet-api-2.5-6.1.14.jar
> servlet-api-3.0.20100224.jar
> stax-api-1.0.1.jar
> stax-api-1.0-2.jar
> ```
> As far as I see it, these are transitive dependencies.
> Teh problem is that what classes are available at runtime depends on which 
> jar is loaded first by the classloader. This is somewhat unpredictable and 
> can lead to weird errors in the components that have these dependencies. 
> ---------------- Imported from GitHub ----------------
> Url: https://github.com/stratosphere/stratosphere/issues/386
> Created by: [StephanEwen|https://github.com/StephanEwen]
> Labels: 
> Created at: Thu Jan 09 19:44:07 CET 2014
> State: open



--
This message was sent by Atlassian JIRA
(v6.2#6252)

Reply via email to