-Dmaven.test.skip=true

[Andrew Savory <an...@luminas.co.uk>]

Hi,

I see this far too often whenever I try working with c2.2:

	-Dmaven.test.skip=true

I mean, really .... why do we have tests if we tell everyone to build  
Cocoon 2.2 without them?

So, suggestions:

- we set the build to by default skip tests
- we do not do a release until all tests pass

WDYT?


Thanks,

Andrew.
--
Andrew Savory, Managing Director, Luminas Limited
Tel: +44 (0)870 741 6658  Fax: +44 (0)700 598 1135
Web: http://www.luminas.co.uk/
Sourcesense: http://www.sourcesense.com/

Re: -Dmaven.test.skip=true

[Carsten Ziegeler <cz...@apache.org>]

Alexander Klimetschek wrote:
> Carsten Ziegeler schrieb:
>> Anyways, I think we should try to fix all tests before a release. If we
>> turn them off by default we could also just remove them as noone will
>> care. (Though the tough question is of course if someone cares today)
> 
> As we use Cocoon as "mission critical" part of our application we 
> definitely care for the quality of the framework ;-)
> 
> Updating/improving the test framework would help also when writing tests 
> for your personal cocoon extensions (SourceFactories, Sitemap components 
> etc.). So if there are any hints on what is to do for updating the test 
> framework, we can provide some help.
> 
Great! Now, I can't give you that much details as I haven't looked at
the stuff for some weeks now. But the easiest way is to just run the
tests in core and have a look at why some of them fail :)
I think one of the biggest problems was the current container handling.
We have a container per sitemap and this information is stored in some
thread local variables "somewhere" - and it might be that this
information is not correctly handled in the tests which causes NPEs or
something like that. But it's too long ago to really remember...

HTH
Carsten

-- 
Carsten Ziegeler - Chief Architect
http://www.s-und-n.de
http://www.osoco.org/weblogs/rael/

Re: -Dmaven.test.skip=true

[Alexander Klimetschek <al...@mindquarry.com>]

Carsten Ziegeler schrieb:
> Anyways, I think we should try to fix all tests before a release. If we
> turn them off by default we could also just remove them as noone will
> care. (Though the tough question is of course if someone cares today)

As we use Cocoon as "mission critical" part of our application we 
definitely care for the quality of the framework ;-)

Updating/improving the test framework would help also when writing tests 
for your personal cocoon extensions (SourceFactories, Sitemap components 
etc.). So if there are any hints on what is to do for updating the test 
framework, we can provide some help.

Alex

-- 
Alexander Klimetschek
http://www.mindquarry.com

Re: -Dmaven.test.skip=true

[Carsten Ziegeler <cz...@apache.org>]

Andrew Savory wrote:
> Hi,
> 
> I see this far too often whenever I try working with c2.2:
> 
> 	-Dmaven.test.skip=true
> 
> I mean, really .... why do we have tests if we tell everyone to build  
> Cocoon 2.2 without them?
> 
> So, suggestions:
> 
> - we set the build to by default skip tests
> - we do not do a release until all tests pass
> 
> WDYT?
> 
The problem is that our own test framework which makes testing avalon
components easier is not working 100% after all the redesigns we had
recently. So we need to fix that...

Anyways, I think we should try to fix all tests before a release. If we
turn them off by default we could also just remove them as noone will
care. (Though the tough question is of course if someone cares today)

Carsten

-- 
Carsten Ziegeler - Chief Architect
http://www.s-und-n.de
http://www.osoco.org/weblogs/rael/

Re: -Dmaven.test.skip=true

[Jorg Heymans <jh...@apache.org>]

Andrew Savory wrote:

> I see this far too often whenever I try working with c2.2:
> 
>     -Dmaven.test.skip=true
> 
> I mean, really .... why do we have tests if we tell everyone to build 
> Cocoon 2.2 without them?

To give them the same user experience as the 2.1 build ? :-)

...

We aren't providing snapshot builds and people want to try out the new 
features. This currently means they have to compile themselves. Failing 
unit tests are a hurdle that might stop their attempt to explore 
completely - the huge 'BUILD FAILED' banner at the end doesn't always 
convey the right message.

> So, suggestions:

We should start providing snapshots again and push the archetype as main 
entry point for those wanting to explore.



Jorg