Follow up to the discussion "jclouds use of gson" from a while back.
My original query here [1]. Response below, and updated PR is [2]. regards Geoff [1] https://lists.apache.org/thread.html/r526ecf03b8cb0819ae37167e9c427bfffe71414ecbe84c6ea5e10215%40%3Cnotifications.jclouds.apache.org%3E [2] https://github.com/apache/jclouds/pull/78 ---------- Forwarded message --------- From: gurkerl83 <notificati...@github.com> Date: Sat, 17 Oct 2020 at 19:04 Subject: Re: [apache/jclouds] Replace embedded and repackaged GSON library (#78) To: apache/jclouds <jclo...@noreply.github.com> Cc: Geoff Macartney <geoff.macart...@gmail.com>, Mention < ment...@noreply.github.com> @geomacy <https://github.com/geomacy> My assessment of whether the change is a critical change is as follows. The main area of concern is the way Apache JClouds is used within Apache Brooklyn? At compile time, runtime or both? If the integration is done at compile time, or running tests, there are no changes. Runtime No change during normal JClouds startup. However, if Apache Brooklyn and JClouds are run in an OSGi runtime environment, and Apache Karaf in particular, there may be some kind of incompatibility. I did a little digging in Apache Brooklyn and found references in one of the git submodules, that Apache Karaf-JClouds features are loaded and started. Here is the reference to the corresponding file https://github.com/apache/brooklyn-server/blob/ba7fb9112623872de18bd7b6135fdfc1014228ae/karaf/features/src/main/feature/feature.xml Per se this is no problem, but Apache Karaf-Jclouds project is on a very old state, which does not map dependencies of newer Apache JClouds versions. The stack and libraries used in Apache JClouds change over time; consequently, the feature set provided by Apache Karaf-JClouds becomes incompatible over time. The creation of Apache Karaf features is mostly a manual process; it simply does not keep runtime dependencies up to date. Possible solution When I made the GSON handling changes in June, I created an automatic feature generator and aligned the outdated Apache Karaf-JClouds project with it. If I find the time in one of the coming weeks I will continue the approach and bring it to finalization, a manual adjustment of features especially in a project with extensive dependencies must be done automatically in the future. The detection and resolution of possible incompatibilities must consider the launch of Apache Brooklyn, whether each of the outdated feature set of Apache Karaf-JClouds can be resolved. Thx! — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <https://github.com/apache/jclouds/pull/78#issuecomment-711058485>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ADL42J6KIODGSKBRHZMGU3TSLHMDXANCNFSM4OIMFOOA> .