[
https://issues.apache.org/jira/browse/HADOOP-6315?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12799677#action_12799677
]
Chris Douglas commented on HADOOP-6315:
---------------------------------------
HADOOP-5281 carried over the semantics from the existing code, i.e. returned
BuiltInZlibInflater when the native libs aren't loaded:
{noformat}
- return ZlibFactory.getZlibDecompressorType(conf);
+ return ZlibFactory.isNativeZlibLoaded(conf)
+ ? GzipZlibDecompressor.class
+ : BuiltInZlibInflater.class;
{noformat}
which made the fix incomplete. Unfortunately, extending HADOOP-5281 to support
the standard platform (i.e. writing the corresponding
BuiltInGzipInflater/Deflater passing nowrap == true) isn't possible either,
since the GZIP\*Streams [don't expose constructors taking
Inflaters/Deflaters|http://bugs.sun.com/view_bug.do?bug_id=6282206] (so it's
not possible to write a .gz file with settings available on a Deflater). Given
that, the current patch makes it impossible to reuse them, which is regrettably
correct.
+1
> GzipCodec should not represent BuiltInZlibInflater as decompressorType
> ----------------------------------------------------------------------
>
> Key: HADOOP-6315
> URL: https://issues.apache.org/jira/browse/HADOOP-6315
> Project: Hadoop Common
> Issue Type: Bug
> Components: io
> Reporter: Aaron Kimball
> Assignee: Aaron Kimball
> Attachments: HADOOP-6315.2.patch, HADOOP-6315.3.patch,
> HADOOP-6315.patch
>
>
> It is possible to pollute CodecPool in such a way that Hadoop cannot read
> gzip-compressed data.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
