I would like stick to the actual approach, that each repo must be explicitly set. this avoids too much "magic" in a build script

Including mavenLocal() in your build script can mean that your build script is not portable. A required dependency that is in your local maven repo may not be present in that of another developer, so what works on your machine may not work on anothers. We would discourage the use of mavenLocal in your build script, unless it is required for consuming artifacts locally published by Maven.

As of Milestone 7, Gradle will automatically use artifacts from the Maven local repository, but only if those artifacts could have been downloaded from one of the a declared repositories. So Gradle will check the remote repo, download the SHA1 file, compare it against the files in the local maven repository, and copy the artifact from their if possible.

We feel that this gives the best possible tradeoff between reproducibility and performance. Some remote network traffic is required to verify the previously downloaded artifact, but the benefit is that your build will work irrespective of what's actually present in .m2/repo.

For this reason, I'm closing this issue as 'wont-fix'. Please let me know if this doesn't make sense to you.

Thank you Daz !

Your solution seems to be quite intricate. However it's a great feature!

Thank you.