>This is really a mess, CI gives different failures than locally... What is >going on with that?
That's not really true. If you run the same build than the CI system, which is a standard Maven build, you'll get the same failures. >You cannot support that many versions of guava with API changes. Like the >removal of getHostText and getHost not being in < guava 20. Also seems trying >to support older targets has issues. We don't really support that many versions. We are very clear on what we support and are just a few range of JDK, Guice and Guava versions. You can see it in our [compatibility matrix](https://jclouds.ci.cloudbees.com/job/jclouds-compat/). Anything outside that matrix is not officially supported and you may find issues (like this already known issue when using Java 8 with Guava >= 21). >Not sure how much more time i want to spend on this. I have already fixed my >issues. This seems like a can of works for JClouds. I also do not understand >why the legacy guava and targets. It will cause issues like this as things >move forward. And this is OK. You took your time to open a PR that you thought fixed the issue but then discovered that the issue was bigger than expected. We already knew that and had several discussions on the development mailing list regarding the best way to fix it, because we already went that path and ended up in finding some issues with the compiler. With my comments, (needless to say we highly appreciate all contributions) I just wanted to point out that the actual change was not enough to completely fix the issue and tried to explain how to build jclouds the standard way so you could properly reproduce the failures. >Maven is horrible and this is taking way to long to fix. I have to fix each >class by class. Maven stops after the first failure. In my environment, I see >all issues so I can address all at once. This is taking really long and is a >very inefficient style of development. About to switch over to looking for all >use of Predicate... You might find it useful if you run it with the "fail at end flag": `mvn -fae` >As a project you all need to make some hard decisions. I maintain hundreds, >over a thousand Java packages. This is one of the only with such legacy >issues. The other is horrible lombok.... I would really recommend telling >those with legacy needs to use older jclouds. For any newer releases really >need to move to newer stuff. Or you will continue to be plagued and held back >by odd issues like this. Which take considerable time to resolve. As I said, we know the existing challenges we'll have to face when moving forward and upgrading JDK and dependencies. And we'll make that decision. Not a hard one, but a *community decision*. One/two years ago we could simply not require Java 8 and a modern Guava because many of our downstream users could not upgrade to modern Guava versions or newer JDKs. And we take backwards compatibility very seriously. I think now it is a bit different as Java 8 has been widely adopted and it is time to start planning for which release we will officially support Java 8 and decide which Guava versions we'll support. If we haven't done that before, it's not because "we don't care". Just the opposite. We care about our users and we try to make upgrades as pain-free as possible, even if that means not using cutting-edge libraries and JDKs for a while. >I think I am about done here... The change seems pretty massive, and I feel >like other issues in future. CI failing on overrides, which at some point >would need to be in there. But why it cannot find the Guava implementation of >test. That to me seems like some legacy issue. Yes, the change could be big. But we still don't know. We still don't understand the cause of the root issue (the one I copied in my previous replies and the last one you pasted), and we need to do some research on that. Once that is done and the cause is properly understood and isolated we'll be able to properly determine the scope of the issue, the right fix, and the steps to upgrade to modern JDK and Guava versions. I completely understand you may not have time to go through this, and it is OK. Thanks for the effort you put on this and for taking your time to have a look at it! -- You are receiving this because you are subscribed to this thread. Reply to this email directly or view it on GitHub: https://github.com/jclouds/jclouds/pull/1180#issuecomment-366428204
