Hi On 19.04.18 23:26, Nir Lisker wrote: > So you're adding read edges in the module-info files via "requires". I > didn't do this because these files are shared and we can't have 2 > versions of them.
Yes as a temporary workaround. This way I could launch the JUnit-Tests! > I added the read edges via "--add-reads" in the .classpath files. This > way Eclipse solves its own problems without external changes. However, > since Eclipse modular support is still in progress, it's impossible to > create a valid run configuration with these modules. There are several > bugs being tracked regarding this approach (I can dig them up if you're > interested). Right hence I used the static trickery! If you CC me on the tickets I can bug the JDT-Team a bit to get them sorted out ;-) Most of the problems are anyways because of the logging stuff. I anyways wonder what the proposed setup for JUnit-Tests is in a JPMS world. Should they really be part of the same application-module? > > Once the java.logging dependency is removed we will be able to simplify > our solutions and I'll post my .classpaths. > Right this is the biggest blocker! > Well until someone proofs that those setups can be generated I > think should maintain them. > > > If no one disagrees by the time java.logging is removed I'll file an > issue to update the Eclipse files. > Yes please do so. Can you share yours? As I said with my "requires static" trick I can run JUnit-Tests, ... from within my Eclipse Photon install (install is the wrong term because I'm running inside an Inner-Eclipse with JDT from master check out!) > Dialog: see https://bugs.eclipse.org/bugs/show_bug.cgi?id=532850 > <https://bugs.eclipse.org/bugs/show_bug.cgi?id=532850> > > > I assume it's the "The blank final field dialog may not have > been initialized" error. It doesn't stop the building at least, unlike > the module stuff. > Right. As you can see in the bug-report this is an edge case for compilers. Tom
