Jason Van Zyl wrote:
>
> This is because the Gump descriptor for Turbine has no direct
> correlation to Turbine. It's a data entry job to keep it in sync and I'm
> not interested in that approach. I'm more interested in an integrated
> approach that ensures accuracy from the project side.
>
> This point will be moot once the Gump descriptor is created from the
> Maven project descriptor. They will always be accurate.

As long as the process can be bootstrapped, I'm willing to work with it.
Example: I used to build and execute your jar fetcher process from the tdk
until the dependencies became circular.

>> I must also admit to a bit of scepticism with respect to the claims made to
>> Maven given this state.
>
> Fair enough, again I have to ardently stress that Gump not being able to
> build a project is not an indication of dissolution. The problem is the
> disconnect between the real project information and the information in
> the Gump descriptor.

Trust me, I have been following this closely for some time.  What you
describe only represents a small percentage of the issues identified.

> And this is important because I disagree with you that you will be able
> to stay on top of a list of packages that grows to something to the
> scale of CPAN. There is just no way.

There is no reason to keep the number of contributors constant.

>> By contrast, I know that the Tomcat team(s) run their own builds using
>> their own processes, and rarely see Gump failures.
>
> They more than likely don't change dependencies often, or the layout of
> the project. The Turbine build usually doesn't build because the dep
> information isn't correct not because a project that Turbine relies on
> has changed their contracts. I hope to change this and I'm trying to
> make the Turbine Gump builds work but I am not going to try and manually
> keep track of 10 descriptors.

Perhaps a better example would be Avalon.  Avalon has been hightly dynamic
and hightly responsive.

> All right. We'll start with this then:
>
> 
>http://cvs.apache.org/viewcvs/jakarta-turbine-maven/jakarta-turbine-maven.xml?rev=1.1&content-type=text/vnd.viewcvs-markup
>
> This is produced by Maven, taking a Maven descriptor and converting it
> to a Gump descriptor. So I have a couple issues I'd like to deal with.
>
> 1) How would you like these descriptors to arrive for use in Gump. I
> would prefer not to mail in a patch, I would just like to be able to
> push it somewhere for you to use. Or I could automatically check it in
> to cvs. I haven't tried Tim Endres cvs client lib but I could give it a
> shot. Thoughts?

I would prefer a process that bootstraps - i.e., build maven's
dependencies.  Build maven.  Execute maven.  Incorporate the dependencies.
Continue on.  I realize that some invention would be required to make this
all work, but I am willing to do my half of the bargain.

> 2) The other issue is the the way Gump embeds projects inside one
> another. Currently Maven depends on GNU regexp (we're working on
> removing it) and wanted to know if a project descriptor could be setup
> this instead of embedding the project definition. The same issue is
> present in the Turbine 2.x descriptor (webmacro, freemarker, helma
> xmlrpc) So can I break out the embedded projects and make descriptors
> for them that live on their own? So that they can be reused? I want to
> add support back in for webmacro and freemarker in Turbine 3.x so I
> would rather not define the project in each project. I assume I can just
> break them out.

I don't follow.  Can you provide an example?

> If we can work this out then I will start sending you accurate
> descriptors and we'll change the way Turbine generally behaves with
> Gump.

Cool.

- Sam Ruby


--
To unsubscribe, e-mail:   <mailto:[EMAIL PROTECTED]>
For additional commands, e-mail: <mailto:[EMAIL PROTECTED]>

Reply via email to