Switch to CircleCI

[danielcompton]

Opening early for feedback. One thing that is tricky is that CircleCI doesn't have the concept of allowable failures like Travis does. At the moment Emacs master is failing on cider-jack-in-clj&cljs (timing out?), but 25 and 26 are running successfully: https://circleci.com/workflow-run/dff69ef0-ec14-4c24-8f19-457e4168af37.

Still needed:

• Update hacking_on_cider doc to show how to do the same thing with CircleCI
• Remove/convert the Travis Docker scripts to CircleCI scripts

---

Before submitting the PR make sure the following things have been done (and denote this
by checking the relevant checkboxes):

• The commits are consistent with our contribution guidelines
• You've added tests (if possible) to cover your change(s)
• All tests are passing (make test)
• All code passes the linter (make lint) which is based on elisp-lint and includes
  • byte-compilation, checkdoc, check-declare, packaging metadata, indentation, and trailing whitespace checks.
• You've updated the changelog (if adding/changing user-visible functionality)
• You've updated the user manual (if adding/changing user-visible functionality)

---

See also clojure-emacs/clomacs#16, clojure-emacs/clojure-mode#514, and clojure-emacs/inf-clojure#159.

[bbatsov]

Do we need Lein at all?

[danielcompton]

cider-jack-in-clj&cljs requires lein to be around. See the test failure at https://circleci.com/gh/danielcompton/cider/11 when removing Leiningen from the build.

[bbatsov]

I'll take a look at the test. We don't do any real jack-ins anywhere in the tests, so probably something just needs to be stubbed.

[bbatsov]

Maybe we should have separate steps for the testing and linting, so that we would get nicer reports at the end?

[danielcompton]

The issue with this is that the majority of CI time is spent getting the image booted and into the right state to run the tests. The lint and test phases only take 6s and 3s. Given this is a free service, I wanted to try and minimise resource usage where practicable. If we wanted better errors from CircleCI we could look at producing JUnit XML reports which they can render.

[bbatsov]

Fair enough. My only issue is that the linting produces a lot of (hard to read) output and this makes it hard to find the tests at least for me.

[danielcompton]

I've migrated the docs for running locally, I think this is now ready to be reviewed. Still not quite sure what to do about failures from emacs-master though. There is an open feature request for this, but not sure when it will be addressed.

[bbatsov]

Well, perhaps we can just remove this build or fix the failure there? I don't see any other options. :-)

[bbatsov]

Actually checking the report you've linked it seems to me that output is not bad even now. I'm still a bit confused when navigating Circle's UI, as I was pretty used to Travis's.

I'll see what I can do about that failing test. Weirdly I don't see anything in the report and I also don't see anything special in the test.

[bbatsov]

Thanks for tackling this, @danielcompton! Much appreciated!