[
https://issues.apache.org/jira/browse/HADOOP-14428?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16033870#comment-16033870
]
Aaron Fabbri commented on HADOOP-14428:
---------------------------------------
Ok [~liuml07], I confirmed that your patch fixes the reproduction I mentioned
(mkdir with trailing slash).
One question on your patch: I'm wondering why the existing test code did not
catch this bug? It looks like you added new test cases that also use
unqualified paths. I would guess that the trailing slash gets removed in
S3AFileSystem in qualify(), but only when it is not absolute? So the existing
test did not catch it?
My reproduction uses an absolute path though, so I'm confused.
Anyways, just curious. I am still +1 on the patch assuming you get a clean
Yetus run.
> s3a: mkdir appears to be broken
> -------------------------------
>
> Key: HADOOP-14428
> URL: https://issues.apache.org/jira/browse/HADOOP-14428
> Project: Hadoop Common
> Issue Type: Bug
> Components: fs/s3
> Affects Versions: 3.0.0-alpha2, HADOOP-13345
> Reporter: Aaron Fabbri
> Assignee: Mingliang Liu
> Priority: Blocker
> Attachments: HADOOP-14428.000.patch, HADOOP-14428.001.patch
>
>
> Reproduction is:
> hadoop fs -mkdir s3a://my-bucket/dir/
> hadoop fs -ls s3a://my-bucket/dir/
> ls: `s3a://my-bucket/dir/': No such file or directory
> I believe this is a regression from HADOOP-14255.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
