[
https://issues.apache.org/jira/browse/HADOOP-18175?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17516870#comment-17516870
]
Steve Loughran commented on HADOOP-18175:
-----------------------------------------
code}
[ERROR]
testStatistics(org.apache.hadoop.fs.s3a.fileContext.ITestS3AFileContextStatistics)
Time elapsed: 1.591 s <<< ERROR!
java.io.EOFException: Attempted to seek or read past the end of the file 512
at
org.apache.hadoop.fs.s3a.read.S3InputStream.throwIfInvalidSeek(S3InputStream.java:440)
at
org.apache.hadoop.fs.s3a.read.S3InputStream.seek(S3InputStream.java:235)
at
org.apache.hadoop.fs.s3a.read.S3PrefetchingInputStream.seek(S3PrefetchingInputStream.java:162)
at org.apache.hadoop.fs.FSInputStream.read(FSInputStream.java:85)
at
org.apache.hadoop.fs.FSDataInputStream.read(FSDataInputStream.java:98)
at
org.apache.hadoop.fs.FCStatisticsBaseTest.testStatistics(FCStatisticsBaseTest.java:107)
{code}
it's happening in the seek(oldPos) code in the default implementation of
positioned read, because the previous unpositioned read
{code}
byte[] buf = new byte[blockSize];
int bytesRead = fstr.read(buf, 0, blockSize);
fstr.read(0, buf, 0, blockSize);
{code}
This is quite a good little failure; I'm impressed but who ever managed to
think at this corner case. pos is at EOF, so returning to it is raising the
EOF.
We will probably want seek() to not do the checks, and have read() do the
validation, which is mostly what s3a does , as it only rejects negative offsets.
> test failures with prefetching s3a input stream
> -----------------------------------------------
>
> Key: HADOOP-18175
> URL: https://issues.apache.org/jira/browse/HADOOP-18175
> Project: Hadoop Common
> Issue Type: Sub-task
> Components: fs/s3, test
> Affects Versions: 3.4.0
> Reporter: Steve Loughran
> Priority: Major
>
> identify and fix all test regressions from the prefetching s3a input stream
--
This message was sent by Atlassian Jira
(v8.20.1#820001)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
