Re: Gump, centipede, maven, etc...

[Nicola Ken Barozzi]

[EMAIL PROTECTED] wrote:

Gump, centipede, maven, etc...

Over the last few days, I've gotten a greater percentage of the gump run

to be attempted. And the work that I did would not have been practical without the recent efforts of Stefan and Peter. While this is clearly a

labor of love on my and other's parts, relying us to do this across all ASF projects (even only the Java ones) simply does not scale.

I agree here. One of the problems with the current Gump configuration is that the groups doing development don't feel that the descriptors are 'their problem', as many are in alexandria CVS.

Yup.

That's why I made Centipede use the Gump descriptor, so that also usual builds use it as a reference; developers will become accustomed to it and need to have it ok just to build.

It

At the beginning, I was a bit concerned that Maven rolled out its own descriptor, but now that the gump plugin is able to create pretty good Gump descriptors, and seeing your active committment on it, I'm confident things will go well.

Getting someone from a team to commit the descriptor to their CVS tree and notify their development group when the build fails helps a lot, IMHO.

>

This should also help with the scalability problem, in that its not a small group working for the benefit of the greater community.

Gump needs more community.

Aye.

I believe that the information that a pure bootstrapping approach provides is quite valuable. Here are a sample of current failures:

I agree.

http://cvs.apache.org/builds/gump/2002-11-14/jakarta-turbine-jcs.html

http://cvs.apache.org/builds/gump/2002-11-14/jakarta-turbine-stratum.html

There is a talk on the Turbine list to switch to Avalon-based services ASAP, so it would be cool to see how things would evolve, and how these changes will affect projects depending on it.

Jelly is also important, given the number of projects that depend on it.

http://cvs.apache.org/builds/gump/2002-11-14/commons-jelly.html
http://cvs.apache.org/builds/gump/2002-11-14/jakarta-james.html

http://cvs.apache.org/builds/gump/2002-11-14/krysalis-ruper.html

Last run was ok, and we're adding other projects.


I'd add that I tried to run Gump some time ago, but failed.

So I tried to understand more of it, split it in pieces over the Alexandria CVS, and finally it started churning, but I still had bugs to fix.

Eventually I had learned enough to understand that the division was not really necessary, so I'm back to trying to run it from the current config.

I've recently gotten a HDSL connection at work, so it will be easier to try Gump runs instead of over a 56k connection.

When I get it running, Bob's my uncle and finally I can start patching with understanding. It's hard to debug when the compile-run-patch cycle is over a day 8-)

From what I've been told James hasn't built successfully for around months, due to changes in the Avalon container.

For this information to be available on a more consistent and timely basis, there needs to be a way for the information necessary to do a bootstrapped build to be actively maintained. This information may originate from a maven project descriptor, or from a unified gump/centipede/etc descriptor.

Furthermore, the code that actually runs the build (be it maven or cent) needs to be aware of the bootstrapped versions of dependencies and substitute them in place of the ones that the developer may have declared.

I don't know how to get this dones across all of the various codebases, but I am willing to do my part. Are there maven and/or centipede developers out there who share this goal?

I'm happy to do my part.

--
Nicola Ken Barozzi                   [EMAIL PROTECTED]
            - verba volant, scripta manent -
   (discussions get forgotten, just code remains)
---------------------------------------------------------------------


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