[
https://issues.apache.org/jira/browse/ACCUMULO-1769?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13792664#comment-13792664
]
Sean Busbey commented on ACCUMULO-1769:
---------------------------------------
Presuming ACCUMULO-1658 gets fixed, is the general case still a concern? As a
part of that we can update the README to call out the need to have a
configuration that doesn't point to your production information.
How conservative do we want to be here? I was under the impression that the
functional tests were already defensive in their generated configurations,
since giving them access to any part of the production configuration is
dangerous.
The primary motivator for reading classpaths from accumulo-site.xml is so that
we can cut down on how many places we have to maintain things like support jars
for hadoop client. An alternative would be to switch to relying on `hadoop
classpath`. The downside to this is that it's overly broad, given that for
Hadoop 2 we only declare hadoop-client as needed.
> Auto test should use classpath from accumulo-site.xml if possible
> -----------------------------------------------------------------
>
> Key: ACCUMULO-1769
> URL: https://issues.apache.org/jira/browse/ACCUMULO-1769
> Project: Accumulo
> Issue Type: Improvement
> Components: test
> Affects Versions: 1.4.4, 1.5.0
> Reporter: Bill Havanki
> Priority: Minor
> Labels: test
> Attachments: ACCUMULO-1769.patch
>
>
> The TestUtils.py code under test/system/auto contains its own definition of
> the general classpath. It should try to read and use the classpath that is
> defined in accumulo-site.xml instead, just as it locates Zookeeper hosts.
> This ensures that tests use the same classpath as the surrounding cluster.
--
This message was sent by Atlassian JIRA
(v6.1#6144)
