hadoop-hdfs-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Jakob Homan (JIRA)" <j...@apache.org>
Subject [jira] [Updated] (HDFS-8939) Test(S)WebHdfsFileContextMainOperations failing on branch-2
Date Mon, 24 Aug 2015 23:56:46 GMT

     [ https://issues.apache.org/jira/browse/HDFS-8939?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]

Jakob Homan updated HDFS-8939:
------------------------------
    Attachment: HDFS-8939-branch-2.002.patch

Branch 2 has a test for these custom properties, which of course fails with their removal.
 Updated the patch to remove the test, but am getting a bit concerned about the backwards
compatibility of this change.  The concept of changing defaults is of course wonky and is
correct to remove; not sure I'm convinced it's safe to remove it as part of branch-2, however...

> Test(S)WebHdfsFileContextMainOperations failing on branch-2
> -----------------------------------------------------------
>
>                 Key: HDFS-8939
>                 URL: https://issues.apache.org/jira/browse/HDFS-8939
>             Project: Hadoop HDFS
>          Issue Type: Bug
>          Components: webhdfs
>    Affects Versions: 2.8.0
>            Reporter: Jakob Homan
>            Assignee: Jakob Homan
>             Fix For: 2.8.0
>
>         Attachments: HDFS-8939-branch-2.001.patch, HDFS-8939-branch-2.002.patch
>
>
> After HDFS-8180, TestWebHdfsFileContextMainOperations and TestSWebHdfsFileContextMainOperations
are failing with runtime NPEs while instantiating the wrapped WebHDFSFileSystems because {{getDefaultPort}}
is trying to access a conf that was never provided.  In the constructor both both WebHdfs
and SWebhdfs the underlying (S)WebHdfsFileSystems are instantiated in the constructor and
never have a chance to have their {{setConf}} methods called:
> {code}  SWebHdfs(URI theUri, Configuration conf)
>       throws IOException, URISyntaxException {
>     super(theUri, new SWebHdfsFileSystem(), conf, SCHEME, false);
>   }r{code}
> The test passes on trunk because HDFS-5321 removed the call to the Configuration instance
as part of {{getDefaultPort}}.  HDFS-5321 was applied to branch-2 but reverted in HDFS-6632,
so there's a bit of a difference in how branch-2 versus trunk handles default values (branch-2
pulls them from configs if specified, trunk just returns the hard-coded value from the constants
file).
> I've fixed this behave like trunk and return just the hard-coded value, which causes
the test to pass.
>   There is no WebHdfsFileSystem that takes a Config, which would be another way to fix
this.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Mime
View raw message