Re: [jgit-dev] Modifying/Testing JGit? Error from testBadCleanFilter tes

[Matthias Sohn <matthias.sohn@xxxxxxxxx>]

On Tue, May 24, 2016 at 10:11 PM, Jeanderson <jeandersonbc@xxxxxxxxx> wrote:

Hi Roy,

I've got the same issue. You will be surprised if you rerun the tests a given number of times: This particular test passes and fails randomly.

This is a Flaky test. I remember finding other tests that for some reason are non-deterministic:

org.eclipse.jgit.api.CommitCommandTest#commitSubmoduleUpdate

I can reproduce this on Mac, it fails every 20..50 test runs when run repeatedly.

I pushed a fix for review

https://git.eclipse.org/r/#/c/73539

 

org.eclipse.jgit.api.AddCommandTest#testBadCleanFilter

org.eclipse.jgit.api.AddCommandTest#testCleanFilterReturning12

org.eclipse.jgit.api.AddCommandTest#testBadCleanFilter2

I can't reproduce this on Mac even if repeating these tests 1000 times, though on Hudson running on Linux

they fail sometimes. I pushed a change to help finding the root cause:

https://git.eclipse.org/r/#/c/73540/

 

I wonder if Hudson has disapproved any Gerrit review because of this randomness (which may not be so frequent but do exists).

I used "mvn clean install" and also "mvn test". I don't know if I'm missing something although I haven't find much information about testing JGit itself. If there are other additional steps, perhaps they should have been documented somewhere.

this happened a few times, we reran the tests in this case

-Matthias