[
https://issues.apache.org/jira/browse/HADOOP-2512?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12555824#action_12555824
]
Amar Kamat commented on HADOOP-2512:
------------------------------------
What I meant was (rewriting a part of the Raghu's patch)
{code}
exitCode = process.waitFor();
+ try {
+ // make sure that the error thread exits
+ errThread.join();
+ } catch (InterruptedException ie) {
+ LOG.warn("Interrupted while reading the error stream", ie);
+ }
+ completed = true;
if (exitCode != 0) {
- if (errMsg.length() == 0) {
- errMsg.append("Command exit with status code " + exitCode);
- }
+ errMsg.insert(0,"Command exit with status code " + exitCode + "\n");
throw new IOException(errMsg.toString());
}
- completed = true;
{code}
Something on similar lines. Does this make sense?
> error stream handling in Shell executor
> ----------------------------------------
>
> Key: HADOOP-2512
> URL: https://issues.apache.org/jira/browse/HADOOP-2512
> Project: Hadoop
> Issue Type: Bug
> Components: util
> Affects Versions: 0.16.0
> Reporter: Raghu Angadi
> Assignee: Raghu Angadi
> Priority: Blocker
> Fix For: 0.16.0
>
> Attachments: HADOOP-2512.patch
>
>
> Fix a couple of issues while handling error stream in Shell (added in
> HADOOP-2344) :
> # fix typo in {{System.getProperty("line.seperator")}}, currently it adds
> "null" instead of "\n".
> # completed is not set to {{true}} when a process exits with an error.
> # In normal error case, it reads errMsg (to create IOException) before
> waiting for errThread to complete, which results in in consistent error
> message. I will attach a patch.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
