Testing the generated package code.
Just run
just test-all
And Bob's your uncle!
An npm dependency can also be a project folder. So this it what we use in the tester project. Then we run tests against that dependency. This dependency is an extract of the tgz package that would be pushed to the registry, so the tests are run on something that is very close the the thing that our users will use.
Another option would be to use the package, the .tgz file as a dependency. This works, but! In CI a clean install fails when this package is missing and this package can sometimes be missing as it needs to be build first. Also, npm pack does not seems to be deterministic. The tgz packages that are created locally are different from the ones created in CI. The contents is exactly the same but the hashes of the packages are different. Also the sizes do not exactly match. The cause seems to come from the compression, the TAR streams are bit identical. This difference makes a clean install break because the hashes do not match.
When running npm install without the referenced folder being present, we get no error! This is kind of a surprise, because you'd expect a dependency to break when it is not present. This is not happening because of the package-lock file. The dependency is described there enough to build the dependencies. Even if the target, the directory, does not exist.