Tests required

[paulirish]

We're lacking tests for a few areas. Let's use this ticket to track them.

• diff-highlight path is resolved on a manual install
• diff-highlight path is resolved when it's already in the PATH
• diff-highlight path is resolved with a more complex symlink setup (homebrew install or nvm'd npm install)
• works without extra gnu-sed dependency (requires fix: Sed error: bad flag in substitute command #49, Making diff-so-fancy work on OS X out of the box #55)
• different meta color doesn't break the diff header formatting (requires fix: Changing the "meta" color reverts the divider to the default style #12)

Please add any others you think we'll need! Edit or leave a comment.

[cht8687]

so after the merge the task works without extra gnu-sed dependency can be marked as done right?

[stevemao]

Still missing comprehensive tests for it.
In fact, the tests would already fail if the default sed is not working.

[OJFord]

@paulirish Can you clarify the difference between resolving d-h when in $PATH, and the following item?

I'm not familiar with nvm'd npm installs, but Homebrew will complain (via brew doctor) if /usr/local/bin isn't in $PATH, which contains symlinks to diff-so-fancy, diff-highlight, etc. - so isn't just checking it works from $PATH sufficient? (c.f. #74, working on my brew install.)

[paulirish]

@OJFord I was just differentiating between the cases of 1 symlink involved vs 2+ symlinks. #36 also mentions the oddness around this.

[OJFord]

Oh I see, I hadn't realised this caused problems.

[paulirish]

@OJFord well the new readlink functionality at the top of the file loops to resolve, because asking for readlink foo can return another symlink. so its a test for that coverage.