Hi Sander, you're right, it's a bug, --add-modules should not be necessary.
Rémi ----- Mail original ----- > De: "Sander Mak" <sander....@luminis.eu> > À: "jigsaw-dev" <jigsaw-dev@openjdk.java.net> > Envoyé: Vendredi 13 Janvier 2017 11:39:10 > Objet: #CompileTimeDependencies and module resolution > When you have a module A with an optional dependency on B (`requires static > B`), > module B is not resolved from the module path unless `--add-modules B` is > provided at startup (tested with jigsaw-b151). In [1] this is mentioned as a > detail, and in fact it is phrased as a question there. Is this still an open > question? Because in my mind, expressing an optional dependency *and* having > to > add the module as root module to get it resolved seems a bit over the top (and > typically you'd also have to add it as dependency in your build tool of choice > to get it on the module path). > > "If it's on the module path, use it, otherwise, ignore it" is intuitively more > appealing to me for optional dependencies. It also more closely resembles the > way things currently work on the classpath with many frameworks. > > For linking, I can see why adding optional modules explicitly is better. > Still, > I think not having to do this at run-time (just like service provider modules > are resolved at run-time but not at link-time) makes more sense. > > > Regards, > Sander > > > [1] > http://openjdk.java.net/projects/jigsaw/spec/issues/#CompileTimeDependences
