-
Notifications
You must be signed in to change notification settings - Fork 2
Description
I'm testing using the platform's *.aggr which I've modified to add more rules for completeness of the mapping, including correctness for all dependencies:
The above selected maven mapping specifies to map any bundle to $key$ where key must be present as a property in the p2 metadata of the installable unit for the rule to be applicable. This normal default rule kicks in when that's not the case.
A new tab in the Analyzer editor presents the information about the POM induced for each installable unit. Here we see that package imports are resolved to one or more installable units (two in the case selected below) and those are also mapped to dependencies:
The Analyzer editor analyzes each generated POM to ensure that the group/artifact information can be resolved to its maven metadata either remotely or locally, e.g.,
https://repo1.maven.org/maven2/org/eclipse/jdt/org.eclipse.jdt.apt.pluggable.core/maven-metadata.xml
It also tests that each dependency's version range resolves to an existing artifact version.
It's not clear whether adding support for java.package requirements will result in bogus or undesirable dependencies. In the example above. I think the dependency on ecg is undesirable (and definitely unnecessary), but I don't know if it's problematic. If you look at the most recent published version you can see that the dependencies for package imports were not present previously:
In other cases duplicates seem correct, e.g., as in this case involving split packages:
My sense though is that we need some general mechanism to control this new behavior or even to turn off this new behavior...



