ping @tianon @LK4D4 @tiborvass @icecrime

Thanks @unclejack !

Pinging testify lovers @aluzzardi @vieux: what do you think is the right tradeoff between asserts and ifs (e.g. unclejack@9f10d27#diff-ca97ecb464e792b1086fca2ee8af23f8R39)

I am +1 to use all packages of testify if that will be more convenient then current situation. Also probably multiple suites makes sense, for example for daemon tests.

So I started looking a bit at testify to see what it does, and I feel it is very wrong and wouldn't use it for Docker as it is. I submitted stretchr/testify#128 to fix this kind of unexpected behavior:

• assert.Equal(t, 2.2, 2) == true and assert.Equal(t, 2, 2.2) == false
• assert.Equal(t, 'x', "x") == true and assert.Equal(t, "x", 'x') == false

So it's mostly NOT LGTM for me unless we get these fixed.

@aluzzardi ^^

@icecrime That's OK, we can avoid using assert right now and do that later.

Being able to use the suites to have setup and teardown for the tests is extremely useful.

This is labelled as "design review" but I don't seen any changes to UI or public API. What am I missing?

Well I wasn't sure if deciding if we want to use testify counts as "design review" or "code review"

Sorry @unclejack, I'm closing this until we can agree on whether or not to use testify. I'm personally +1 for this, and would love to use both the test suites as well as the assert functions. If there are some weirdnesses that still need to be fixed on it, let's get those PRs merged upstream.

Please continue discussion below.