[
https://issues.apache.org/jira/browse/HADOOP-13045?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15254541#comment-15254541
]
Akira AJISAKA commented on HADOOP-13045:
----------------------------------------
Thanks Allen for the comments.
bq. In any case, HADOOP_USER_CLASSPATH should be used to do what you're trying
to do.
I got it.
bq. What if we turned the current .hadooprc support into .hadoopenv and add a
new .hadooprc hook that gets called after the initialization is done such that
functions work?
I thought that .hadooprc was where to use unix shell API, but it's not. Where
can/should we use the API? If there are some places to use the API, I'll use
there. I'd like to avoid a new .something, which maybe misunderstanding. If
there are no places to use that functions, I'm okay to introduce a new
.something.
> hadoop_add_classpath is not working in .hadooprc
> ------------------------------------------------
>
> Key: HADOOP-13045
> URL: https://issues.apache.org/jira/browse/HADOOP-13045
> Project: Hadoop Common
> Issue Type: Bug
> Reporter: Akira AJISAKA
>
> {{hadoop_basic_function}} resets {{CLASSPATH}} after {{.hadooprc}} is called.
> {noformat}
> $ hadoop --debug version
> (snip)
> DEBUG: Applying the user's .hadooprc
> DEBUG: Initial CLASSPATH=/root/hadoop-tools-0.1-SNAPSHOT.jar
> DEBUG: Initialize CLASSPATH
> DEBUG: Rejected colonpath(JAVA_LIBRARY_PATH): /usr/local/hadoop/build/native
> DEBUG: Rejected colonpath(JAVA_LIBRARY_PATH): /usr/local/hadoop/lib/native
> DEBUG: Initial CLASSPATH=/usr/local/hadoop/share/hadoop/common/lib/*
> {noformat}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
