Convert tests to RSpec

[twe4ked]

No description provided.

[twe4ked]

$ git diff master..rspec --stat
 .travis.yml          |    2 +
 Gemfile              |    4 +
 Gemfile.lock         |   35 +
 Makefile             |    2 +-
 spec/fresh_spec.rb   | 2095 ++++++++++++++++++++++++++++++++++++++++++++++++++
 spec/spec_helper.rb  |  214 ++++++
 spec/support/bin/git |   65 ++
 test/fresh_test.sh   | 2065 -------------------------------------------------
 test/support/shunit2 | 1048 -------------------------
 test/test_helper.sh  |  153 ----
 10 files changed, 2416 insertions(+), 3267 deletions(-)

Wow. Finished. Let me know what you think @jasoncodes.

[star-szr]

This looks really nice! I'm looking into adding test coverage to another shell script: https://github.com/virtualhost/virtualhost.sh.

Could you share your main reason(s) for switching to RSpec? I'm guessing one would be that shunit2 looks to be abandoned but I'm also wondering if you've found the experience to be a positive one. I've always loved that fresh has test coverage so I've been following this PR with great interest :)

[twe4ked]

@cottser There are a few reasons for changing. The first is simply for organisation, describe/context blocks mean we can group related tests together. I also find it easier to add helper methods to make the test readability better. Having access to thing like active support's strip_heredoc method is fantastic.

Another reason is the error messages we get. I find coloured diffs much easier to parse.

I also find RSpec's --color option really useful.

Ruby's system call is great and using active support to capture stdout/stderr worked well.

I'd recommend it if you're going to have more than a dozen tests. It was nice having the tests in the same language as the actual code but I think this is going to be easier going forward.

[star-szr]

Awesome, thank you very much for the response @twe4ked! I will be digging into this code in the near future to learn as much as I can.

[twe4ked]

No worries, run_fresh is probably the most interesting method. It's pretty horrible but it works quite well. The strange order of the assertions are so we get the best error messages. For instance if we expect stdout to be empty we check that before checking stderr and the exit code so we can see what we got in stdout. Not sure if that made sense.

Another two options to look at are bats and shpec. We decided against shpec because of this open issue: rylnd/shpec#15. I don't think we ever really looked into Bats properly.

[twe4ked]

@cottser Looks like there is some work happening on shpec that might be interesting. If they have an end function it may mean they can add the cleanup code.

rylnd/shpec#43

[star-szr]

Nice, thanks @twe4ked :)