Fog::Mock.reset is not resetting image (glance) mock data.

[zeroDivisible]

So, I'm not entirely sure if this a bug in fog-openstack - or an incorrect assumption made in fog-core (I'd say a bit of both).

The problem is that if I understand correctly, if one wishes to reset the mocks when testing, this can be achieved by calling

Fog::Mock.reset

The reset method is implemented in fog-core:

def self.reset
  mocked_services = []
  Fog.constants.map do |x|
    next if Fog.autoload?(x)
    x_const = Fog.const_get(x)
    x_const.respond_to?(:constants) && x_const.constants.map do |y|
      next if x_const.autoload?(y)
      y_const = x_const.const_get(y)
      y_const.respond_to?(:constants) && y_const.constants.map do |z|
        next if y_const.autoload?(z)
        mocked_services << y_const.const_get(z) if z.to_sym == :Mock
      end
    end
  end

  mocked_services.each do |mocked_service|
    next unless mocked_service.respond_to?(:reset)
    mocked_service.reset
  end
end

So: get constants under Fog (i.e. :Image), get constants under Fog::Image (i.e. :OpenStack) and then get from Fog::Image::OpenStack - if one of them is called :Mock, make it a candidate for resetting.

The problem is that there are different classes for different versions of Glance API: i.e. Fog::Image::OpenStack::V1::Mock - so the Mock class in the hierarchy is under V1, not under OpenStack, where fog-core is searching for it.

Until this issue is fixed, a workaround is to i.e. call:

Fog::Mock.reset
Fog::Image::OpenStack::V1::Mock.reset

to manually reset those mocks - however, I think that a proper solution would need to be added. I think that the same issue will apply to other implementations of versioned APIs (i.e. Keystone), though I didn't check this.