[icedtea-web] Broken unit-tests

[Jiri Vanek]

On 04/24/2012 05:28 PM, Omair Majid wrote:
> On 04/23/2012 10:02 AM, Jiri Vanek wrote:
>> I really do not see an connection between your changeset,  the failure
>> and the fix you are suggesting.
>

I wrote it quite badly. My point was, that I was not sure why your changeset needed resources.jar. 
There was no  need for it as I read it.
However I analysed the stacktrace wrongly -  and so I have not seen the "deeper need" for resources.jar.

However patch is in and have solved the issue. And the folk saying "Omair is (nearly) always right" 
force me to dig what was so needed form resources.jar  in such s simple (and understandable) patch.
(any comment if you will have enough time will be appreciated O:)
 >
> Well, you are probably aware that 'make check' was working before that
> changeset was pushed to the repo, right?  So Something in that changeset
> is breaking the test runner. So there are two way to fix it:
>
> 1. Fix the test so it doesn't break the test runner.
>
> 2. Fix the test runner so it can handle this problematic test.
>
> Your patch (from what I can tell) takes approach 2. The fix for
> resources.jar takes approach 1 (and makes the test pass).

True :)

>
> It would be good to have fixes for both 1 _and_ 2, so that no future
> test can break the runner and the current test passes.
>
Sure:)

>> If you have tried and it is working for you, I think you can push. As it
>> can be useful to have resources.jar on classapth anyway.
>
> I will push it in a bit.

Thanx for pushing this!

>
>> However - to fix this issue is better to  catch the proper exception in
>> XmlOutputLIstener (which is probably consumed somehow and is causing the
>> class lock itself in error state).
>
> Oh, yes. The test runner needs to be fixed too. If you want to post an
> isolated patch for just this, I will try and review it.
>

The patch will not be as isolated as you wish, because I have already merged it with one more 
changeset (where I got into issue with test mechanism null). It was nothing else then one  null 
cpomarsion and one more try catch. I will try to separate it during review in all ways.

Best regards, J.