A few weeks back, Dan committed the Eclipse publishing plug-in into the SVN trunk. It
really is a fantastic and nicely-done bit of software. He did almost all of the
development on a separate branch, and at the end merged the code into trunk. Since this
was our first non-Maven code, I suggested that he commit it into the 'tools'
directory, thinking that we'd be able to integrate the Eclipse build into the Maven
build system.
But after thinking about it for a while, building the Eclipse plugins (including the
update site) from within our build system is a pretty complicated affair, both from a
mechanism perspective but also from a dependency perspective. The Eclipse plugin requires
the released JARs for 'dna-common' and 'dna-web-jcr-rest-client' to be
bundled into the plugin update site, but those JARs really have to be built against the
tagged release. That means that the Eclipse plugin code would have to be tagged with the
rest of the code as part of the normal release, but the Eclipse update site would have to
be reproduced afterward. Any problems with the Eclipse plugin or update site, and the
whole codebase would have to be re-tagged and re-released. Very messy.
My latest idea is that the Eclipse plugins just shouldn't even be in the
http://anonsvn.jboss.org/repos/dna/trunk/ at all, and to instead place them in a separate
area of SVN, like
http://anonsvn.jboss.org/repos/dna/eclipse/trunk (along with the
corresponding 'branches' and 'tags' sibling directories). All of the
Eclipse plugin code would move into there, where they can remain non-Maven and
Eclipse-specific PDE projects. We can then release the Maven-built part of DNA as usual,
and then bundle and release the Eclipse plugin code.
(At some point in the future, the DNA Eclipse plugin code might even move into the JBoss
Tools superproject. So it certainly doesn't hurt to break it out now.)
Anybody have thoughts about this?
Best regards,
Randall