hadoop-common-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Harsh J (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HADOOP-11512) Use getTrimmedStrings when reading serialization keys
Date Mon, 02 Feb 2015 02:32:35 GMT

    [ https://issues.apache.org/jira/browse/HADOOP-11512?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14300828#comment-14300828

Harsh J commented on HADOOP-11512:

Thanks again - the (1) change looks good, but the test still will not fail without the fix,
making it incomplete. We need (2) done too to achieve it.

In literal terms, from the SerializationFactory object constructed at the end of your test,
invoke its {{getSerializer(Text.class)}} or so, and assert that the returned object is not
'null'. This will exercise the serialization lookup and complete the test.

The current behaviour of constructing SerializationFactory with a bad config is that it still
passes omitting the bad classes it finds in the configs (choosing to log them as WARNs than
to throw an exception).

Let me know if this makes sense!

> Use getTrimmedStrings when reading serialization keys
> -----------------------------------------------------
>                 Key: HADOOP-11512
>                 URL: https://issues.apache.org/jira/browse/HADOOP-11512
>             Project: Hadoop Common
>          Issue Type: Bug
>          Components: conf
>    Affects Versions: 2.6.0
>            Reporter: Harsh J
>            Assignee: Ryan P
>            Priority: Minor
>         Attachments: HADOOP-11512.patch, HADOOP-11512.patch
> In the file {{hadoop-common-project/hadoop-common/src/main/java/org/apache/hadoop/io/serializer/SerializationFactory.java}},
we grab the IO_SERIALIZATIONS_KEY config as Configuration#getStrings(…) which does not trim
the input. This could cause confusing user issues if someone manually overrides the key in
the XML files/Configuration object without using the dynamic approach.
> The call should instead use Configuration#getTrimmedStrings(…), so the whitespace is
trimmed before the class names are searched on the classpath.

This message was sent by Atlassian JIRA

View raw message