feat(api): inject release metadata into application

[bacongobbler]

closes #1057

rebased off of #1078 for the _build_env_vars tests, see e88c52f for the new commit.

[deis-bot]

@helgi, @kmala and @Joshua-Anderson are potential reviewers of this pull request based on my analysis of git blame information. Thanks @bacongobbler!

[helgi]

Is this what the ticket that was created asked for? Thought there was a workflow prefix and all?

[bacongobbler]

Given deis/builder#418 and Heroku both use SOURCE_VERSION it makes sense to use that convention. I can add the other environment variables here, one sec

[helgi]

Ah is that the only one that was not prefixed? Sounds good

[kmala]

any specific reason for this change or just a clean up

[bacongobbler]

this is the rebased commit from #1078, but essentially this is cleanup from dead code. get_port used to return a NoneType if the port didn't exist; now it just raises a DeisException or a numerical value so we don't need this check any more

[codecov-io]

Current coverage is 87.12% (diff: 100%)

Merging #1080 into master will increase coverage by <.01%

@@             master      #1080   diff @@
==========================================
  Files            42         42          
  Lines          3601       3603     +2   
  Methods           0          0          
  Messages          0          0          
  Branches        609        610     +1   
==========================================
+ Hits           3137       3139     +2   
  Misses          307        307          
  Partials        157        157          

Powered by Codecov. Last update ad182b2...096948f

[kmala]

should we check for sha and add as environment as it is available even for dockerfile apps

[bacongobbler]

Sure, that's not a bad idea.

[bacongobbler]

FYI this is done now

[ultimateboy]

Seems like we should add these new vars to the test here: https://github.com/deis/controller/blob/master/rootfs/api/tests/test_app.py#L585

[bacongobbler]

They have been added! See the changes to test_app.py in this PR. :)

[helgi]

Does this need to go through strftime? Just curious since we set the format already for DRF so I'd expect it to translate it for us when we do str() on it, but maybe not?

[bacongobbler]

I think it does because we do it elsewhere in the codebase. DRF just handles it in the serializer, I think.

[helgi]

Did we consider the ambiguousness that was brought up in #1057 and what was the final decision? I see we only cover the SOURCE_VERSION story when sha is available

[bacongobbler]

Did we consider the ambiguousness that was brought up in #1057 and what was the final decision? I see we only cover the SOURCE_VERSION story when sha is available

With this pr we now cover WORKFLOW_RELEASE_SUMMARY, WORKFLOW_RELEASE_CREATED_AT and SOURCE_VERSION which is WORKFLOW_RELEASE_HASH. I didn't bother with WORKFLOW_RELEASE_HASH because SOURCE_VERSION already has that data.

[felixbuenemann]

@bacongobbler Thanks for implementing this!

So for docker image deploys I would have to set SOURCE_VERSION when building the image, right?

[bacongobbler]

Yep!

[bacongobbler]

Wait no. If you're using Dockerfile or buildpack deploys the environment variable will be injected into your environment.

[felixbuenemann]

@bacongobbler Docker image != Dockerfile =) So with "docker image deploys" I was talking about deis pull deploys.

[bacongobbler]

Okay then yes if you're using deis pull you'll have to inject that into the environment yourself because we don't have any git metadata to derive the sha from in deis pull. The only information we have is the image and the release summary.