Tracepoint Update Site
My Tracepoint plugin for Eclipse begins with an update site. Before any real code is written, I want to know that the results will be published in a timely mannor.
I’m working off of the How to Keep Up to Date article at eclipse.org.
This isn’t dovetailing with the new structure I’ve eastablished for my builds. Eclipse creates many projects, with long qualified names, in parallel in the Eclipse IDE.
I don’t have as many projects. Instead, I’m dividing my projects into artifacts.
A project is something I can talk about, like Tracepoint, XStrategy, Relay, or Momento, while and artifact is some subset of functionality, delivered as a JAR, like xstrategy-sax, relay-mail, momento-xslt. This works well with Ivy.
There is a parallel in the Eclipse builds, between features and plugins, but on the filesystem, they are all next to each other, no heirarchy. In my layout, an artifact is in a subdirectory under the project. In Eclipse, the project is one project, but on the command line, Ant compiles each target separately.
I’m not sure if there is some form of auto-discovery for Eclipse when checking out of source control, if it can determine which dependencies are necessary and pull them out of the respository. If so, I’d to set up my projects so as not to defeat this feature.
I don’t like using the fully qualified names to name directories. I still spend a lot of time in ViM when working with Java. Using fully qualified names, com.agtrz.tracepoint, for directory names is hard on me. I don’t like tapping the tab completion twice in bash. Lazy me. I’m able to name my projects differently than the directories in which they reside, but I don’t know this will hold up during plugin development.
Yoiks! Publishing this plugin is not going to be easy. I’m not going to run FTP on my server, athough WebDAV is an option. Yeah, why not? It’s running for Subversion, and I see I can turn it on for any directory. Okay. The mod_dav documetation, for when I get around to this.
I thought I was going to have a automated build on the server when I was done with this. One that build and published on the server. Reading the article Build and Test Automation for plug-ins and features, it’s obvious that this will not happen any time soon.
Apparently, Eclipse uses Ant to build it’s plugins. The obfuscated Ant used to extend the build in the article is par for the course. (It reassures me that that the notion of a Pure Ant build for a complicated project is a boondoggle.) Ultimately, the authors of the article extend the build to use Subversion, by cutting out targets, and replacing them with shell scripts.
My update site for Tracepoint is complete. You can install the do-nothing default tracepoint plugin from it’s update site, at http://engrm.com/eclipse/tracepoint/.
The last step after building was…
scp ./* alan@engrm.com:/www/eclipse/tracepoint
… but, as noted, I’ll move to WebDAV, and use that to publish.
As far as that goes, I’ll be looking for a WebDAV client for Java, like maybe an Ant task based on Jakarta Slide.