[
https://issues.apache.org/jira/browse/HADOOP-9665?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13721176#comment-13721176
]
Chris Nauroth commented on HADOOP-9665:
---------------------------------------
I intend to merge this patch to branch-1-win later today. The branch-1 patch
applies cleanly to branch-1-win.
> BlockDecompressorStream#decompress will throw EOFException instead of return
> -1 when EOF
> ----------------------------------------------------------------------------------------
>
> Key: HADOOP-9665
> URL: https://issues.apache.org/jira/browse/HADOOP-9665
> Project: Hadoop Common
> Issue Type: Bug
> Affects Versions: 1.1.2, 2.1.0-beta, 2.3.0
> Reporter: Zhijie Shen
> Assignee: Zhijie Shen
> Priority: Critical
> Fix For: 2.1.0-beta, 1.2.1
>
> Attachments: HADOOP-9665.1.patch, HADOOP-9665.2.patch,
> HADOOP-9665-branch-1.1.patch
>
>
> BlockDecompressorStream#decompress ultimately calls rawReadInt, which will
> throw EOFException instead of return -1 when encountering end of a stream.
> Then, decompress will be called by read. However, InputStream#read is
> supposed to return -1 instead of throwing EOFException to indicate the end of
> a stream. This explains why in LineReader,
> {code}
> if (bufferPosn >= bufferLength) {
> startPosn = bufferPosn = 0;
> if (prevCharCR)
> ++bytesConsumed; //account for CR from previous read
> bufferLength = in.read(buffer);
> if (bufferLength <= 0)
> break; // EOF
> }
> {code}
> -1 is checked instead of catching EOFException.
> Now the problem will occur with SnappyCodec. If an input file is compressed
> with SnappyCodec, it needs to be decompressed through BlockDecompressorStream
> when it is read. Then, if it empty, EOFException will been thrown from
> rawReadInt and break LineReader.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
