[
https://issues.apache.org/jira/browse/HDFS-8885?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14711935#comment-14711935
]
Chris Nauroth commented on HDFS-8885:
-------------------------------------
Hello [~srevanka]. Thank you for posting the patch. Here are a few notes.
# A {{ByteRangeInputStream}} starts out in {{SEEK}} status, which indicates the
underlying HTTP stream needs to be opened. If a caller calls {{available}}
before any {{read}} calls, then the result won't be correct, because it has not
yet determined the file length. I recommend calling {{getInputStream}} first,
which lazily opens the underlying HTTP stream if necessary.
# A similar situation occurs if a caller calls {{seek}} immediately before
{{available}}. Again, calling {{getInputStream}} first would handle this case
correctly.
# It's typical for {{available}} to throw an {{IOException}} if called after
the stream has been closed. Once again, a call to {{getInputStream}} would
cover this case, because that method is already implemented to throw an
{{IOException}} if the stream is in {{CLOSED}} state.
# Can you please look into adding unit tests in {{TestByteRangeInputStream}}?
> ByteRangeInputStream used in webhdfs does not override available()
> ------------------------------------------------------------------
>
> Key: HDFS-8885
> URL: https://issues.apache.org/jira/browse/HDFS-8885
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: webhdfs
> Affects Versions: 2.7.1
> Reporter: Shradha Revankar
> Assignee: Shradha Revankar
> Priority: Minor
> Attachments: HDFS-8885.000.patch, HDFS-8885.001.patch
>
>
> ByteRangeInputStream has to override the method {{available()}} , the super
> class implementation returns 0. Clients using the method {{available()}} to
> rely on available bytes information will end up with errors during reads if
> WebhdfsFileSystem is used.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
