FWIW I had talked with Jason Vanzyl yesterday here at Apachecon and asked him about how they would
do the conversion. It was a short conversation but his comment was that reorganizing the tree wasnt
overly critical but it makes sense at the right time.
One of the suggestions he had was to make a list of things that need to be done and then get them
done one at a time. In his experience trying to move to a full M2 environment from M1 is best done
incrementally and not all at once. As Jason pointed out there may be some loss of functionality but
that should be weighed against productivity.
I have only been following this thread as e-mails come in but have not expended the time you guys
have so forgive me if I ask obvious questions.
Is there a set of steps that we need to accomplish to get the build working
under M2?
Is there a priority order to them?
What are the fit and finish things we can do after the conversion is complete to allow people that
working on Geronimo the ability to be as productive as possible?
I know Prassad and Anita have been working on this very hard and I've seen David Jencks and others
pulling this together. I think it would be good to outline the actions and help everyone understand
what needs to be done to help get this important work completed.
Also, can we start a new thread like "M2 Issues and Actions" :)
Jason Dillon wrote:
If we do move things around in trunk, will it make merging changes
made in the 1.1 branch more difficult?
Most likely... not sure how well SVN tracks changes and then merges back
from branches after bits have been moved about.
<soapbox>
I know that Perforce would be able to handle these types of merges...
</soapbox>
Basically it will mean that files will need to be merged one by one
explicitly, not using the recursive mechanism.
If so, how important is it to move things now and would there be a
better time to do it, e.g. when 1.1.1 is released?
Well, I believe it is important... moderately important. We eventually
need to bite the bullet and make these changes, which will cause some
additional work when merging due to the way that SVN works. It is work
that must be done at some time, and I think that the sooner the better.
Its not just the organization of the module's files... but also the
organization of the modules themselves. IMO we MUST do both to take the
most advantage out of the new m2 build system.
I can't stress enough that the current layout was designed around m1.
m2 is quite different with respect to the rules that apply when organizing.
I'm not sure that delaying these changes will making anything easier.
It may reduce work by 5-10% in the short-term but will probably increase
work in the long run if we keep delaying to keep merges simple. If we
really want to keep merges simple, then we should use Perforce, which
actually handles incremental merging and can handle any arbitrary
branching and handle full integration history when merging back off of
branches of branches. Sorry, back on the soapbox again... but really
IMo there is no better SCM than Perforce for large projects that need
advanced branching and merging.
--jason
