I have a similar problem. I have the following unit test:

FreeStyleProject project = createFreeStyleProject( "drmemory" );

final InputStream in = getClass().getClassLoader().getResourceAsStream( "results.txt" );

/* To get the real workspace and copy the results.txt */
project.getBuildersList().add(new TestBuilder() {
    public boolean perform(AbstractBuild<?, ?> build, Launcher launcher,
        BuildListener listener) throws InterruptedException, IOException {
        build.getWorkspace().child("drmemory").child( "1" ).child( "results.txt" ).copyFrom( in );
        return true;
    }
});

Builder builder = new DrMemoryBuilder( "", "", "drmemory" );
project.getBuildersList().add( builder );
		
DrMemoryPublisher publisher = new DrMemoryPublisher();
project.getPublishersList().add( publisher );
		
FreeStyleBuild b = project.scheduleBuild2( 0, new Cause.UserIdCause() ).get();

which results in a "Failed to clean up temp dirs" error, because the publisher copies the file "drmemory/1/results.txt" to the builds folder and it is unable to delete that file.

Status update:

• testRotation(hudson.util.io.ReopenableRotatingFileOutputStreamTest): fixed by pull request jenkinsci/jenkins#582; to be merged
• testCopyTo2(hudson.FilePathTest): the test has been renamed to testNoFileLeakInCopyTo; I have not been able to reproduce this error yet, probably because I have a 32-bit JDK
• testRemoting(hudson.util.ProcessTreeTest): still failing, although I get a slightly different error message: Failed to open process error=87 at .\envvar-cmdline.cpp:53
• testIsSymlink(hudson.UtilTest): fixed (see JENKINS-14641)

So, I'm down to 1 failing test, but I don't have a clue how to even start looking for a fix for this one. I can't even find the referenced C++ source file, envvar-cmdline.cpp. Does anybody have any hints/ideas?

Erik

This mailing list thread seems to think that process id's 0 and 4 should be filtered? http://jenkins.361315.n4.nabble.com/Re-hudson-util-ProcessTreeTest-test-error-td2269355.html

It looks like process id 0 is the system idle process and process id 4 is another system process, so filtering those two should be ok I think.

Agreed, filtering 0 and 4 should not cause any problems. However, it also doesn't consistently fix the test. It seems to depend heavily on what kind of processes are running, or even what kind of mood Windows is in. For instance, today I get error=87 for PID=0, and filtering out 0 fixes the test. However, two days ago, I got error=5 for PID 2200, belonging to svchost.exe, which was linked to 5 services. Today, those same 5 services are also running (this time with PID 2188), but they do not make the test fail.

I just found that the problem with ProcessTreeTest is also described in: https://issues.jenkins-ci.org/browse/JENKINS-8614
Looks like it's been around for a while (reported January 2011).

Since all the other test errors reported in this issue have been solved, this remaining problem is in fact a duplicate of 8614. I therefore suggest that this issue be closed.

as suggested in the comments, resolving as duplicate of JENKINS-8614