Community
Participate
Working Groups
the omplementation of FeatureRootAdvice#isApplicable() implicitly relies on feature ID from feature.xml == maven POM artifactId. However tycho does not enforce this. Probably we should not rely on equality here but rather use a mapping. We should have the mapping info since we also create MavenPropertiesAdvice which contains the IU -> GAV coordinates mapping.
Beware that tycho-versions-plugin assumes Maven artifactId to match Eclipse project id. Not sure if there are other places that make the same assumption, but I wonder if there is a real need to allow such flexibility.
(In reply to comment #1) > Beware that tycho-versions-plugin assumes Maven artifactId to match Eclipse > project id. Not sure if there are other places that make the same assumption, > but I wonder if there is a real need to allow such flexibility. so we need to decide whether we require Maven artifactId == OSGI artifact id or not. If we require it, we should make it explicit and validate it.
(In reply to comment #2) > If we require it, we should make it explicit and validate it. I am +1 on adding and validating this requirement. Too much flexibility can be a problem too.
(In reply to comment #3) > I am +1 on adding and validating this requirement. Too much flexibility can be > a problem too. I agree. While thinking about it again I just see one potential problem: I guess there are many existing eclipse projects with the same ID for a plugin and a feature. If we validate Maven artifactId == OSGI artifact id, then they have to either 1. use different maven groupIds for feature and plugin - or - 2. change feature or plugin ID (which in general is an incompatible change)
I agree that it is good practice to have matching artifactIds, but I don't think that we should force an incompatible change on our users. Therefore I think we will need to allow artifactId != Eclipse ID. The only option I see is to enforce artifactId == p2 ID. The p2 IDs anyway need to be unique. Not sure people will like the "org.example.feature.feature.group" IDs for eclipse-feature modules though...
Just a thought -- Tycho does not have to support every single possible use case and usage scenario. There are always going to be cases Tycho does not or cannot support.
I am now also in favour of enforcing artifactId == Eclipse ID. With the option to have different groupIds (e.g. abc and abc.features), there is still enough flexibility for people who need to keep IDs for compatibility reasons. I've created bug 353384 for the proposed change. This would solve this issue.
With bug 353384 implemented, the assumption feature ID == artifactId is valid. Therefore the problem of missing root files can no longer occur.
