[
https://issues.apache.org/jira/browse/HADOOP-3918?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12621618#action_12621618
]
Pete Wyckoff commented on HADOOP-3918:
--------------------------------------
The new patch includes a unit test when the classpath is just set to foo.jar to
ensure it recognizes there is no hadoop-blah.jar and exits with exit code 4.
The proper tests I added hadoop.jar to the classpath to spoof it. (since in
test mode, it is just using build/classes and not the jar).
> Improve fuse-dfs error message when classpath not set or set incorrectly
> ------------------------------------------------------------------------
>
> Key: HADOOP-3918
> URL: https://issues.apache.org/jira/browse/HADOOP-3918
> Project: Hadoop Core
> Issue Type: Improvement
> Components: contrib/fuse-dfs
> Reporter: Pete Wyckoff
> Assignee: Pete Wyckoff
> Priority: Minor
> Attachments: patch0.txt, patch1.txt
>
>
> Currently libhdfs does an exit(1) when it cannot create a hadoop object. fuse
> does not like this and instead fuse_dfs should return -EIO and log to syslog
> and stderr a message. Also if it is obvious somehting is wrong, like the
> CLASSPATH is null or does not contain something of the form
> hadoop-[a-zA-Z-].jar immediately print an error and exit
> Note that it is tough to debug this issue because when fuse-dfs is not in
> debug mode, it is running as a fuse module and stderr is redirected, so all
> one gets is "software abort" when libhdfs exits. So, this will greatly
> improve things since fuse-dfs can print a proper error and the user can
> quickly see what the problem is.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
