Add support for coveralls.io coverage reports

[maul-esel]

OK, it's true, those badges in the README might be a little obsession of mine. Nevertheless, it occurs more often than it should that some behaviour in the jekyll code is accidentally broken, yet no test goes 💥 like it should - because there is no test for that particular behaviour. In most cases it's detected before merge, but there are also cases that go into the next release and then 💩 goes down.

So, while it is not necessary to test every single line and a coverage of 100,000000 % is surely not necessary, some visible report on it can probably help to improve jekyll's (already awesome) testing suite. This PR adds support for coverage analysis during Travis-CI builds and loading them up to coveralls.io (provided @mojombo signs in there with his github account).

See https://coveralls.io/r/maul-esel/jekyll for the current reports of this branch.

[coveralls]

Changes Unknown when pulling aa10f8c on maul-esel:coveralls into * on mojombo:master*.

[maul-esel]

Huh cool, didn't even know it does that. One more pro argument.

[maul-esel]

Funny, just did this and now: Test coverage Travis-CI + CodeCilmate. Need to check this out.

[parkr]

So much cruft with the RUBY_VERSION and the ENVs! Argh! Not sure how I feel about this yet. :/

[maul-esel]

Me neither. Let's check out that CodeClimate stuff first, I think that could be integrated much better.

[maul-esel]

OK, it seems CodeClimate test coverage only works if you have a CodeClimate account 💰. So probably not worth it 😭.

As to all the ENV: looking at it again, we could possibly reduce that quite a bit: always run coverage, but only send it if ENV["TRAVIS"]. The slowdown is too small to notice anyway. The RUBY_VERSION however seems non-optional, because it simply doesn't work for 1.8.7.

[coveralls]

Changes Unknown when pulling b26327f on maul-esel:coveralls into * on mojombo:master*.

[maul-esel]

So, I removed the ENV conditions as far as possible.

However, update on CodeClimate: they're working on support for open-source repos, though it might be some time (reference) and they're also working on the ability to combine coverage results from cucumber and unit tests, which jekyll needs.

I really think that on a long-term basis, using CodeClimate is better because having both their features combined next to each other can surely help a lot. Also, it wouldn't require using another service. So IMO the only question left is if we

1. want to use coveralls as an intermediate solution ➡️ merge
2. or wait for CodeClimate ➡️ close
3. or don't want it at all ➡️ close

[parkr]

lol ❤️

[parkr]

I'd be OK with this! @mattr-?