failing network acl tests on travis.

[geemus]

@drcapulet - this travis failure (https://travis-ci.org/fog/fog/jobs/15987405#L1068) seems likely to be related to your recent ACL work, could you take a look?

[drcapulet]

Yeap! This is that transient error I hadn't been able to reproduce, but I think I may have found the source, I'll submit a PR once I've got a fix.

[fcheung]

I think i sort of fixed this in 7940b30. I think the core issue is a test ordering one: create a vpc creates some default network acls, but destroying the vpc doesn't clear those up. For some reason the order that the tests run on travis is different to what I get locally, but running pretty much any test that creates a vpc before the network acl tests seemed to reproduce the failure (for ease of reproduction I just pasted the tests from rds/subnet_groups_tests at the top of network_acl_tests.rb)

I've gotten the tests to pass by resetting the fog mock data at the start of the test, but perhaps the mock for destroying vpcs should clear up the network acls too? Not sure whether that reflects what happens in real life.

[drcapulet]

Hmm, I know we already clean up the default network ACL - https://github.com/fog/fog/blob/master/lib/fog/aws/requests/compute/delete_vpc.rb#L39-41.

I (think) I've traced it down to a subtle issue with the creation of subnets - I'll hopefully have a PR by this evening.

[geemus]

@drcapulet - do you think this should be sorted out now? Just didn't want to close it if you thought there still might be lingering issues.

[drcapulet]

Yeap! Once @fcheung found that set of tests that consistently caused the tests to failure, it became a lot easier once I noticed my subnets.create(...) call wasn't giving me back the object it created.

[geemus]

@drcapulet Great, thanks for the update. Closing this.