Github user steveloughran commented on the pull request:
https://github.com/apache/spark/pull/9306#issuecomment-163222684
Moving to Hadoop 0.90
[HADOOP-9623](https://issues.apache.org/jira/browse/HADOOP-9623) was what could
be described as "an accidental disaster"'; the patch swallowed exceptions
"which should never happen", resulting in
[HADOOP-10589](https://issues.apache.org/jira/browse/HADOOP-10589); a seek(0)
on a 0-byte file NPE-ing. (trivia: It was fixed by probably the only piece of
co-recursive code in core hadoop)
One issue with 0.90 is that the `close()` call on an input stream reads
_all remaining bytes on the resource_
[HADOOP-12376](https://issues.apache.org/jira/browse/HADOOP-12376). This hurts:
moving up to 0.94 may fix it. From the hadoop core perspective, the move to
0.90 broke enough things that we are scared to go near the s3n code again; all
future work is in s3a.
To summarise then: this may break s3n if not shaded, but you should be
encouraging people to use s3a on Hadoop 2.7+ anyway
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
