Adam Spiers

> Totally agree here after having to deal with apply_role, and exactly what I was thinking of doing shortly, extracting apply_role into a minimum of 4 methods (as those are...

The CI failure is a standard haproxy failure which has been happening for a while. I doubt the CI even exercises this code path.

@tboerger Excellent point - I forgot about that since I was thinking of the files as recipes not Ruby files. Will fix!

@tboerger BTW I found [another reason to prefer underscores](https://github.com/sethvargo/chefspec): > WARNING Cookbooks with dashes (hyphens) are difficult to test with ChefSpec because of how Chef classifies objects. We recommend naming...

Oh, that's talking about cookbooks not recipes. Never mind :-)

For the record, I found a way to automatically change all `.rb` files affected by this PR to have underscores instead of hyphens: ``` shell base_commit=`git upstream` # commit on...

Rebased because it looks like the gate failed due to the PR being based on `master` from before the change regarding digest authentication.

And now the gate fails for some weird reason which looks unrelated :-(

@vuntz Rebased. I also locally backported to 3.0 and tested on my c19 which is currently running SOC6 - seemed to work fine, but I didn't test on SOC7 yet.

@tboerger I ran `rubocop -c .hound.ruby.yaml` against the files changed by this PR, and it didn't generate any of the `Style/DotPosition` violations Hound is complaining about - any ideas why?