[
https://jira.codehaus.org/browse/SUREFIRE-817?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=310522#comment-310522
]
Todd Lipcon commented on SUREFIRE-817:
--------------------------------------
I spent some time looking into this, and it appears that it was fixed in
Surefire 2.12.3, probably by one of the following commits:
{code}
commit 30f9991b6645ebb1520fdeb70addffe5c22a3938
Author: Kristian Rosenvold <krosenv...@apache.org>
Date: Mon Aug 20 17:02:21 2012 +0000
o Removed last remains of CDR
git-svn-id: https://svn.apache.org/repos/asf/maven/surefire/trunk@1375109
13f79535-47bb-0310-9956-ffa450edef68
commit f8cb5724b82cd5faec37f1d37f3fdc8dc2bb6643
Author: Kristian Rosenvold <krosenv...@apache.org>
Date: Thu Aug 16 16:38:34 2012 +0000
o Refactored all the duplicated shared state out of the reporters
git-svn-id: https://svn.apache.org/repos/asf/maven/surefire/trunk@1373909
13f79535-47bb-0310-9956-ffa450edef68
commit 22643d3b4c6ac6c2c3f0d48586924b3d8b3194f3
Author: Kristian Rosenvold <krosenv...@apache.org>
Date: Tue Aug 14 14:29:32 2012 +0000
o Worked on simplifying the reporting stuff
git-svn-id: https://svn.apache.org/repos/asf/maven/surefire/trunk@1372903
13f79535-47bb-0310-9956-ffa450edef68
{code}
Given that, I'm inclined to resolve this JIRA. Agreed?
> JUnit 4.7 test output is always buffered, lost if forked process exits
> abnormally
> ---------------------------------------------------------------------------------
>
> Key: SUREFIRE-817
> URL: https://jira.codehaus.org/browse/SUREFIRE-817
> Project: Maven Surefire
> Issue Type: Bug
> Components: Junit 4.7+ (parallel) support
> Affects Versions: 2.11
> Reporter: Todd Lipcon
>
> The junit47 provider and above support multi-threaded test execution, and
> thus interpose a buffering layer (ConcurrentReporterManager) in between the
> test output and the actual stderr/stdout. This is ostensibly to allow the
> multiple threads' stderr and stdout to be demuxed nicely when the suite
> completes. But, if the JVM exits abnormally (eg due to a segfault or a
> System.exit() call), no output is generated. This is problematic since it's
> very hard to debug the test failure!
> In my opinion, the buffering layer should only be interposed _when parallel
> running is enabled_. For the non-parallel case, there's no need to buffer the
> output. A simple test case is to write a JUnit test which prints a line of
> output to stderr and then calls System.exit(1). The output doesn't show up
> anywhere.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators:
https://jira.codehaus.org/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira
