Bug Report Checklist

• Have you provided a full/minimal spec to reproduce the issue?
• Have you validated the input using an OpenAPI validator (example)?
• Have you tested with the latest master to confirm the issue still exists?
• Have you searched for related issues/PRs?
• What's the actual output vs expected output?
• [Optional] Sponsorship to speed up the bug fix or feature request (example)

Description

I had some code that used to work on 5.4.0 and now it doesn't. I traced the issue down to some change in the implementation of the usage of oneOf, where previously swagger::OneOf<A,B> was generated, now a new model is generated, but that new model is empty.

https://gist.github.com/mirosval/433308deb22727e3b1102fb58f3a6652

openapi-generator version

Works on 5.4.0 and does not since 6.0.

OpenAPI declaration file content or url

I'll provide a full gist, but shortly these are the interesting bits:

openapi: 3.0.3

...

components:
  schemas:
    Test:
      type: object
      properties:
        something:
          oneOf:
            - type: string
            - type: integer

Before:

    pub something: Option<swagger::OneOf2<String,i32>>,

After:

    pub something: Option<models::TestSomething>,

...

#[derive(Debug, Clone, PartialEq, serde::Serialize, serde::Deserialize, validator::Validate)]
#[cfg_attr(feature = "conversion", derive(frunk::LabelledGeneric))]
pub struct TestSomething {
}

Generation Details

generate-710:
	docker run --rm \
		-v $(PWD):/local openapitools/openapi-generator-cli:v7.1.0 generate \
		-i /local/test.yaml \
		--generator-name rust-server \
		--output /local/openapi \
		--global-property models,modelDocs=false,supportingFiles
	mv openapi/src/models.rs test-710.rs
	rm -rf openapi

Steps to reproduce

See the Gist and then make generate-540 generate-710 generate-latest

Related issues/PRs

This looks like it might be related:

#13538

Suggest a fix

It seems that generating intermediate models for nested properties is a good idea, just that there should probably be some heuristic to detect when the nested type is a oneOf and use a different path.