[RFC][icedtea-web] Test for regression of ServiceManager not setup correctly

[Danesh Dadachanji]

On 19/02/12 01:38 PM, Jiri Vanek wrote:
> On 02/18/2012 12:08 AM, Omair Majid wrote:
>> On 02/17/2012 05:34 PM, Danesh Dadachanji wrote:
>>> Hi,
>>>
>>> Here's a test that checks whether or not ServiceManager is setup
>>> correctly by the time an applet constructor is called. The constructor
>>> uses BasicService and will throw a NPE if it is not setup.
>>>
>>
>> Would it be possible to extend this test to check that services are
>> available in init(), start(), stop() and destroy() as well?
> +1!
>>
>> There also seem to be some spacing/indentation issues in the patch.
>>
>>> The test is for checking regression against this changeset:
>>> http://icedtea.classpath.org/hg/icedtea-web/rev/221174bcd4ec
>>>
>>> Thank you very much to Jiri for helping me get this through the door!
>>>
>>> +2012-02-17 Danesh Dadachanji<ddadacha at redhat.com>
>>> +
>>> + Adding test for regression of JNLP API accessibility in constructor
>>> + methods of applets.
>>> + * Makefile.am: Added classes.jar to classpath when compiling
>>> jnlp_tests.
>>> + * tests/jnlp_tests/simple/CheckServices/: Tests ServiceManager is
>>> setup
>>> + correctly when called from applet constructors.
>>
>> Please list all the files added or changed.
>>
>> In general, am not sure how good these applet tests might be. We know
>> that applets run under javaws use different code paths from applets run
>> under a browser. Still it's better to have a few tests than none.

I agree that this isn't the most ideal way to test applets but the good 
thing is that once there is a way to test plugin-run applets, we don't 
have to make a new reproducer, we can use the same applet/JNLP later on 
with jnlp_href. The other thing is that applets run from javaws did not 
have ServiceManager setup either by the time the constructor was called 
so it definitely does cover half the problems that can occur.

> I have played with this test and is testing what it should test. => Very
> good test :)

Thanks to you it's up and running! I think what Omair meant was more 
that the idea behind the changeset we're testing about was for applets 
run by the plugin. The test is still valid for those run with javaws but 
it doesn't guarantee plugin-run applets will succeed as well. I'd say it 
takes us half way there though!

> There is still issue with headless appelts, but I'm going to solve ths
> at least by setting DISPALY to tests cron job.
> Also applet-testing-framework will be here soon I hope O:)
>

Yes this would be most ideal. Fortunately, this test is simple enough 
that don't even need to modify it to run with the plugin! =)

Cheers,
Danesh