Test::Harness in Core

[Andy Armstrong]

On 9 Jun 2009, at 13:53, David Golden wrote:
> On Tue, Jun 9, 2009 at 7:50 AM, Andy Armstrong<andy at hexten.net> wrote:
>> On 8 Jun 2009, at 21:38, Steve Purkis wrote:
>> I believe we can still sort out the plugin system without breaking
>> backwards compatibility. App::Prove's public interface is pretty
>> benign - I don't see anything there that'd have to be deprecated to
>> support a better plugin interface.
>>
>> But certainly let's do it! :)
>
> But please do it in Test::Harness.  That's where you need the plugin
> capability to let a simple Test::Harness live in core and let people
> extend it in their own tests.
>
> The fact that so much effort was put into App::Prove and not
> Test::Harness is one reason why this problem still exists.

Agreed that TH needs better extensibility - but in this context I was  
thinking specifically of prove-only plugins. For example things like  
prove's state tracking seem too high level to live inside the harness.  
If prove's plugin interface was better the state tracking could live  
outside the Test::Harness distro.

> For example --- answer this quickly and without reading documentation:
> which of the options to prove are supported by HARNESS_OPTIONS?  Don't
> know?  Go look it up.  Was there any well-thought-out reason for it?
> Or just that no one bothered to keep the code in sync?


HARNESS_OPTIONS was added after the fact to make it possible to run  
parallel tests when using the Test::Harness compatibility layer.

-- 
Andy Armstrong, Hexten