Jetty 12.0.x Simplify tests with OpenId by using an exising openid provider and avoid extra not necessary maven modules

[olamy]

• use keycloak as openid provider for testing
• remove extra modules not needed
• checkstyle
• start keycloak only once
• fix openid module for ee10
• fix typo

[lachlan-roberts]

I don't think using keycloak is necessary to merge these modules, as we have an implementation of OpenIdProvider with jetty-core. But I do think its an improvement, now we test with an actual OIDC provider implementation and not just a mockup.

[olamy]

The idea is to remove code we maintain and using proven existing solution.
Why having code to implement an openid provider as some solutions exist already ;)

[olamy]

@lachlan-roberts init of keycloak done via some API usage. remove of this huge json import file.

result of the cleanup

[lachlan-roberts]

I would have preferred use of the jetty HttpClient instead of the gargoylesoftware one.

[olamy]

bear in mind this is using now a real openid provider so there is a redirection to web page for login.
And not sure we really want to spend time parsing the returned HTML to find all the information such: cookie, all ids pass, session etc....
it;s not only an http client but most important a HTML parser and browser like library
it looks the easiest way to mimic a browser.

[lachlan-roberts]

The Jetty client will handle the session cookies, and can follow redirects, and here it is always reading the full content as a string and doing containsString(...) checks on it, so no HTML parsing needed. And then there is a FORM submission which is also easy to do with the Jetty HttpClient.

[lachlan-roberts]

I'm not 100% sure the rules adding new dependencies, but if you're sure we can add these for testing then LGTM.

Although I still prefer to use Jetty HttpClient for the testing if possible.

[olamy]

I'm not 100% sure the rules adding new dependencies, but if you're sure we can add these for testing then LGTM.

I don't understand the problem. It's just a test module not even deployed/distributed so adding dependency doesn't have any impact.

Although I still prefer to use Jetty HttpClient for the testing if possible.

The problem is to extract data from the returned HTML. It's not like if parsing HTML was like parsing structured/well-formed data :)
We can do it and I will have a try as it's probably not too complicated, but I feel like re-inventing the wheel.

[olamy]

@lachlan-roberts using our own Jetty HttpClient done