remove sleep from tests

[phemmer]

This removes time.Sleep() from the majority of the tests. This should alleviate the racyness that causes the CircleCI tests to randomly fail. It also makes the affected tests significantly faster.
The few tests where time.Sleep() wasn't removed generally one of 2 cases. Either doing so would have made the code very complex, or I don't have the ability to test the change.

Required for all PRs:

• CHANGELOG.md updated (we recommend not updating this until the PR has been approved by a maintainer)
• Sign CLA (if not already signed)

Fixes #2382

[phemmer]

Tests are unfortunately failing on the tail input plugin. The github.com/hpcloud/tail package is racy as hell. Am going to try to submit a patch on it to fix some of the more critical issues. But I can't even get that package's own tests to pass on my system (even on a completely unmodified master), so it might take a little work.

[danielnelson]

We could merge everything except the tail changes now, and fix those tests later.

[phemmer]

This solves the race detector issues.

[danielnelson]

@phemmer Getting lots of failures over the last few days from the tail tests, any sense on if this might be fixed by your upstream fix? https://circleci.com/gh/influxdata/telegraf/6404

[phemmer]

Hrm. Looks different. Still a race condition, but I don't think related to my upstream PR. Will look into it later tonight.

…

-------- Original Message -------- From: Daniel Nelson <[email protected]> Sent: March 21, 2017 7:16:27 PM EDT To: influxdata/telegraf <[email protected]> Cc: Patrick Hemmer <[email protected]>, Author <[email protected]> Subject: Re: [influxdata/telegraf] remove sleep from tests (#2555) We could merge everything except the tail changes now, and fix those tests later.

-- You are receiving this because you authored the thread. Reply to this email directly or view it on GitHub: #2555 (comment) -- Sent via mobile

[phemmer]

That library is so incredibly messy. If there were an alternative available, I'd recommend switching. But I find no alternatives.
Anyway, found & fixed the issue, upstream PR here: hpcloud/tail#116

Looking at the history of the project, PRs seem to take months to merge. The last PR took 5 months to merge in, and there were no changes that had to be made during review process.
Given that, I'm not sure how you want to proceed. The telegraf tail test failures aren't just the tests failing. They're legitimate issues. We can add hacks to get the tests to pass, but that kinda defeats the point of the tests, as real world users can experience the issues.

[danielnelson]

I think we will have to use a fork at least until your changes are accepted. Would you rather merge both of your changes onto a single branch on your fork or should I make one on our organization?

[phemmer]

I think I would prefer it be part of influxdata, or an official member of the team :-)

[danielnelson]

Can you open the PRs on https://github.com/influxdata/tail. I really appreciate all the help, you are official to me but perhaps not to the legal team ;)