Possible old/new property confusion for arrays

[joeduffy]

A customer reported a problem when replacing an array value with another. The particular example was with the aws.cloudfront.Distribution object, namely its aliases property. The resource was provisioned with one (string) value, and then a simple update to a different (string) value triggered a strange error that implies maybe Terraform was seeing two values instead of the expected one.

The update was:

      ~ aliases             : [
          - [0]: "old.domain.io"
          + [0]: "old.domain.com"
        ]

And the update led to the following error:

error PU2003: Plan apply failed: rpc error: code = Unknown desc = updating urn:pulumi:cts-prod::cts::aws:cloudfront/distribution:Distribution::cts-prod-cdn: InconsistentQuantities: The specified quantity of Aliases (2) does not match the actual quantity supplied (1).
    status code: 400, request id: c187413f-fee2-11e7-b4ba-21238073c963
Step #48 failed [update]: this failure was catastrophic and the provider cannot guarantee recovery
info: no changes required:
      47 resources unchanged
A catastrophic error occurred; resources states may be unknown
rpc error: code = Unknown desc = updating urn:pulumi:cts-prod::cts::aws:cloudfront/distribution:Distribution::cts-prod-cdn: InconsistentQuantities: The specified quantity of Aliases (2) does not match the actual quantity supplied (1).
    status code: 400, request id: c187413f-fee2-11e7-b4ba-21238073c963
error: update unsuccessful: status failed

It is very possible this was fixed with the many updates to update logic we've made in the past few weeks, since both the SDK and the PPC that the customer is using are quite out of date.

[lukehoban]

Looking at TF AWS provider code - looks like this condition is triggered here: https://github.com/terraform-providers/terraform-provider-aws/blob/master/aws/cloudfront_distribution_configuration_structure.go#L966

[pgavlin]

I believe that this is now fixed after #104.