The only thing that comes to mind ... this early in my day ... is that 
there is always a problem if you assume tests run in a particular order. 
Not sure if you have that (hidden?) assumption in your tests, or not? They 
might work fine on one machine, with one VM, but then have trouble on 
another machine with a different VM because it "picks" a different order. 

I know JUnit 11 (which we also moved to, up from JUnit 10) has some 
built-in help for this, in the form of an Annotation named 
@FixMethodOrder(MethodSorters.NAME_ASCENDING) 

See 
https://github.com/junit-team/junit/blob/master/doc/ReleaseNotes4.11.md

I hope this gives one possible hint ... but, could be something completely 
unrelated. 

Also, Hudson has also been changing its "proxy rules" at a frequent pace, 
the past few weeks, if you have anytests that depend on the network. 
For a while, it even required a proxy to get to "download.eclipse.org", 
only from some machines. I think that's all settling down as of yesterday 
and as I understand it no longer requires a proxy to get to 'downloads'.) 

Hope this helps, a little. 

 




From:   Ed Willink <[email protected]>
To:     Cross project issues <[email protected]>, 
Date:   03/14/2013 03:46 AM
Subject:        [cross-project-issues-dev] Strange Hudson test failures
Sent by:        [email protected]



Hi David

Is there anything more that you can share with us?

Currently my three projects OCL, QVTd and QVTo are all showing 
unexpected test failures. These are not the usual early collapses 
attributable to a bad Hudson day. So my suspicions point to EMF or 
platform. EMF has only had one change against which I've raised a 
critical regression (Bug 403289), but that does not correlate well with 
the actual test failure pattern. I would only expect Bug 403289 to give 
build failures if a 1.5 JRE is present, which it isn't on our Hudson jobs.

On 8-Mar I spent a long time trying to reproduce QVTo test failures that 
appeared after I removed the org.junit4 dependency. I was unsuccessful 
and when I continued on 9-Mar the problems had vanished. Perhaps it was 
Hudson, perhaps it was the platform that went through a significant 
change.

Looking at the dashboard, I see quite a few recent completions with 
amber status. Is anyone else seeing these issues?

[Hudson UI, particularly the Configure-Save response is incredibly slow 
at the moment, indeed Save frequently doesn't refresh at all, but 
Back-Configure confirms whether the change has occurred.]

     Regards

         Ed Willink

_______________________________________________
cross-project-issues-dev mailing list
[email protected]
https://dev.eclipse.org/mailman/listinfo/cross-project-issues-dev


_______________________________________________
cross-project-issues-dev mailing list
[email protected]
https://dev.eclipse.org/mailman/listinfo/cross-project-issues-dev

Reply via email to