[BUG] [JS] call to inexistant method `validateJson()` on nested `oneOf` class

[etshy]

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?
  Tested on latest release, not latest master.
• 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

Json is gnerated from Php codebase usnig nelmio/api-doc-bundle.
On nested classes (properties, not extension), the js client will start calling validateJson on class that doesn't have that method.
It happens on validateJson method that call other class validateJson, but there is a "proxy class" between both.

Here is a repo with the generated js client for my openapi json : https://github.com/etshy/bad-generated-openapi-js-client

You can see in AddThingsCommand.js, validateJSON method, it try to call AddThingsCommandThings.validateJSON() but the method doesn't exists.
It should do ThingsDto.validateJSON() because AddThingsCommandThings is a "proxy" to ThingDto.

Noting that the generated client is good when using allOf, but nelmio bundle change all allOf to oneOf from v4.13 to follow a change on zircote/swagger-php : nelmio/NelmioApiDocBundle#2156

openapi-generator version

7.0.1 and 7.7.0 using javascript generator

OpenAPI declaration file content or url

Json

Gist

Generation Details

java -jar openapi-generator-cli-7.7.0.jar generate -i ./minimalopenapi.json -g javascript -o ./path/to/

with minimalopenapi.json the gist I linked

Steps to reproduce

Generate js client with provided openapi json

Suggest a fix

When using oneOf and there is a sort of "proxy" class, guess the real class/dto and call validateJson on it instead of the proxy.

[EmilePerron]

For anyone who's still running into this issue, updating to the newly released v7.13 has fixed this on my end.

Specifically, this PR seems to have fixed the issue: #21043

[wing328]

thanks for confirming the fix

[EmilePerron]

As it turns out, it seems the issue remains in some cases.

Is it possible that an edge case has been missed in #21043?
In my case, it happens in the type of a nullable collection that is nested 2-3 levels deep in the properties of some other objects.

[Mattias-Sehlstedt]

Hi @EmilePerron, I did not consider anything more complex than the first-level case (I had an issue with that in a personal project), so it could definitely be that some corner case(s) have been missed.

Could you share some scenario(s)? You could for example share them by forking the repo and then adding an additional test here, so that any one that might implement it know what is your expected result.