Fix bug

[chanseokoh]

Credential itself is an HttpRequestInitializer.

[codecov-io]

Codecov Report

Merging #40 into master will increase coverage by 3.58%.

@@            Coverage Diff             @@
##           master      #40      +/-   ##
==========================================
+ Coverage   74.88%   78.47%   +3.58%     
==========================================
  Files          10       10              
  Lines         223      223              
  Branches       20       20              
==========================================
+ Hits          167      175       +8     
+ Misses         50       42       -8     
  Partials        6        6

Impacted Files	Coverage Δ
...om/google/cloud/tools/ide/login/OAuth2Wrapper.java	0% <ø> (ø)
...google/cloud/tools/ide/login/GoogleLoginState.java	75% <100%> (+4.12%)	:white_check_mark:

---

Continue to review full report at Codecov.

Legend - Click here to learn more Δ = absolute <relative> (impact), ø = not affected, ? = missing data Powered by Codecov. Last update 8a74722...915a6c4. Read the comment docs.

[chanseokoh]

Credential.initialize() is the right one that will put auth info to the HTTP header section (or so, I think). It seems that Credential.getRequsetInitializer() returns null for our Credential objects. I will look into it further.

As for testing, let me try, but this could be a bit challenging. I may have to further contort the code that was already sharded to enable mocking these network things. (The whole point of UserInfoService was to mock these.)

[chanseokoh]

I think I can write a test taegeting only the returned chained initializer. Will add that.

[elharo]

Is the bug in a recent change, not yet pushed? Otherwise I'm left wondering how the current code works.

[chanseokoh]

This new code is what I merged yesterday. The code we are using currently is unrelated.

[chanseokoh]

I believe I found a right way to do this. Chaining initializers and overriding it in OAuth2 works, but now I think it wasn't the right way.

[chanseokoh]

I believe I found a right way to do this.

Nope, too early to have said that.

[chanseokoh]

The UserInfoService has been significantly simplified. In fact, it's now genuinely a shim wrapper for Oauth2, so I was able to change the name to OAuth2Wrapper.

The refactoring allowed me to add tests that concretely verify whether the credential's initializer is called and the timeouts are set.

[chanseokoh]

@elharo PTAL