lint/cclint separately from make test

[jbergstroem]

I've seen a job in jenkins fail once or twice because the lint at the end of testing fails. It'd be great if we could create a bot that first passes linting before passing it on to the "real" tests. Saving cpu cycles, quicker feedback to user, etc.

We already have make lint, but since it assumes iojs is available, this bot might have to be a bit "special". Also, I'm not sure how to set up jenkins this way.

It would also work nice as a step in the github test notification feedback, but that's a later problem

[rvagg]

let's set up a separate job just for linting, add a few new small machines with iojs on them, then add that new job in to the multi-pr job(s):

I think we can add that to the top of the build steps and make it block until that one passes.

[rmg]

👍 for using linting as a gate for running full tests.

[jbergstroem]

Sounds like a plan, then. If no one objects, I'll setup a machine on DO shortly.

[jbergstroem]

We now have two linters. Each run seems to be around ~40 secs (on tree's that are relatively similar and lives in block cache). A cold tree run is about 1min 10sec. The vm's are run in New York and San Fransisco for redundancy. The final step is to have this project block the main project and merge nodejs/node#1965.

[jbergstroem]

Rolled into production a week ago. Closing.