Honor async.error settings from api.yaml in Terraform providers

[melinath]

It looks like the operation error options that you can set in api.yaml don't actually get used anywhere - for example: https://github.com/GoogleCloudPlatform/magic-modules/blob/065f680d36c0a653e9d8542257d60663f96c15c5/mmv1/products/datastream/api.yaml#L40

it looks like the actually operation error just gets wrapped in CommonOpError which just always uses e.Message:

terraform-provider-google/google/common_operation.go

Line 18 in 5800a6d

return fmt.Sprintf("Error code %v, message: %s", e.Code, e.Message)

This breaks down if the error message is in a different place - for example:

...
"error": {
    "code": 9,
    "details": [
      {
        "@type": "type.googleapis.com/google.rpc.ErrorInfo",
        "reason": "POSTGRES_CONNECTION_ERROR",
        "domain": "datastream.googleapis.com",
        "metadata": {
          "message": "An unexpected database connection error occurred while trying to connect to the data source. Make sure that the connection profile configuration is correct, or contact Google Support.",
          "uuid": "de7cb468-4c56-49ec-a32b-09a396f2ecf0",
          "time": "2022-08-30T17:37:43.192113Z",
          "originalMessage": "timeout expired\n"
        }
      }
    ]
  }
...

This ends up getting displayed as:

Error: Error waiting to create ConnectionProfile: Error waiting for Creating ConnectionProfile: Error code 9, message: 

Expected Behavior

We should be honoring the async.error settings so that autogenerated resources can compensate for different operation error structures.

[melinath]

The workaround is to turn off autogen_async and handwrite the operations code. There are currently ~13 handwritten operations files.

[rileykarson]

Based on the number of wrong / untrusted async blogs in MM and the infrequency of operations deviating significantly, we'd rather handwrite exceptional cases for now.

Terraform's templates only use a limited number of fields from the operations.

[melinath]

Closing as a duplicate of #17098 - we should just remove the configuration for now and then do an actual intentional implementation later if needed.

[github-actions]

I'm going to lock this issue because it has been closed for 30 days ⏳. This helps our maintainers find and focus on the active issues.
If you have found a problem that seems similar to this, please open a new issue and complete the issue template so we can capture all the details necessary to investigate further.