@slnode test please

To explain: this change is moving the two recently added test from top-level "end-to-end" block inside nested "empty-server template" block, where the sandbox app required by these two test cases is properly scaffolded by before/beforeEach hooks.

/to @jannyHou @0candy please review
/cc @richardpringle @loay

The changes LGTM as you're just moving the test cases. However, CI is failing on that end-to-end empty-server template test so it doesn't look like it is fixing the issue.

@slnode test please

I have published loopback-boot@2.20.0 to npmjs.org, that seems to have fixed the problem on my machine.

My guess is that npm install run from within the test suite uses the public npm registry, not the staging. @rmg is it something we can fix in our CI build tool? I would expect npm install executed by a test to use the same registry as npm install used to install the dependencies before running the tests.

@slnode test please

@0candy @jannyHou the remaining test failures seems to be intermittent. For example, cis-jenkins tests passed on all platforms, while pr-builder failed on Node v6 with errors that seem unrelated:

Test Result (5 failures / -3)
end-to-end autoupdate.end-to-end autoupdate before
end-to-end discovery.end-to-end discovery before
end-to-end testConnection.end-to-end testConnection before
end-to-end start/stop/restart.end-to-end start/stop/restart before
Middleware Middleware.create(def, cb).should be able to create multiple entries

(see http://ci.strongloop.com/job/loopback-workspace/label=amazon-6/4378/)

I am going to re-run CI again to see if we manage to get a green build. Even if not, I am proposing to merge this pull request, as it's a big step towards getting CI green.

@slnode test please

@bajtos LGTM and I think we can merge it now. The failing ones are downstream for apic and apim.
loopback-workspace also fails due to apic/apim failures.

Landed, thank you @0candy and @0candy for the review.