Build improvements #3126
Description
The build should be able to detect situations in which the client needs manual work on top of what is automatically generated. I can think of two specific cases:
- When a new namespace is generated, the namespace module lacks imports, so it fails until someone manually adds the dependencies. This is fine, but then this module needs to be imported into the client, and if this manual step is missed nothing fails. We should add a test that checks that all namespaces can be imported.
- The integration tests use the ES Docker image. There is nothing that ensures that the correct version of the image is used in each development branch. For example, when the 9.2 branch was created, we forgot to update the build script to use 9.2-SNAPSHOT for the tests, so it was testing against 9.1-SNAPSHOT until very close to the release date, and we missed some failures because of this. We should come up with an automatic way to decide which image to use, or if that is not possible, to check that the version selected matches the branch version.