Seemingly dangling type reference to aRN:ARN

[t0yv0]

Hello!

• Vote on this issue by adding a 👍 reaction
• To contribute a fix for this issue, leave a comment (and link to your pull request, if you've opened one already)

Issue details

Currently schema.json contains the following construcs:

      "$ref": "#/types/aws:index/aRN:ARN",

But there is no corresponding entry in .types, so this is a dangling reference. In addition these constructs happen to be used in addition to "type": "string" which seems suspect.

Steps to reproduce

  jq '.resources["aws:datasync/efsLocation:EfsLocation"]' provider/cmd/pulumi-resource-aws/schema.json

  ...
  
  "inputProperties": {
    "efsFileSystemArn": {
      "type": "string",
      "$ref": "#/types/aws:index/aRN:ARN",
      "description": "Amazon Resource Name (ARN) of EFS File System.\n"
    },

Expected: no dangling ref, either string or ref
Actual: string and dangling ref

[stack72]

@t0yv0 this was added by @pgavlin in #991

via the following code:

			"aws_datasync_location_efs": {
				Tok: awsResource(datasyncMod, "EfsLocation"),
				Fields: map[string]*tfbridge.SchemaInfo{
					"efs_file_system_arn": {Type: awsTypeDefaultFile(awsMod, "ARN")},
				},
			},

Notice awsTypeDefaultFile which does the following:

func awsTypeDefaultFile(mod string, typ string) tokens.Type {
	fn := string(unicode.ToLower(rune(typ[0]))) + typ[1:]
	return awsType(mod, fn, typ)
}

This means we could have been carrying this for quite a while - or maybe this is the intended behaviour?

thoughts?

[t0yv0]

Thanks for looking that up! Yes, it might be intended. I do not think it is doing any real harm now just odd corner cases here and there. I stumbled on it when working on programgen for $fn.Output forms, but eventually concluded it's a broader unrelated issue with dangling imports being generated:

pulumi/pulumi#8324

Then I got a workaround in my PR so I can proceed.

For I just would like to clarify and document intent behind this aRN:ARN. I would also like to check why pulumi schema check accepts this. It seems like schema should not validate, and dangling references should be made prominent.

[mikhailshilkov]

This is largely by design. Broader discussion in #2565 - so I'll close this issue.