Bug - Latest Graph metadata changes causes Generation reports errors using

[nikithauc]

The latest changes microsoftgraph/msgraph-metadata@0e2b708 cause bugs in generation.

[zengin]

@rkodev, @fey101. It looks like the snippet generation errors are due to the last Open API update. The error is not specific to the snippet generation scenario.

[nikithauc]

@zengin Should this issue be closed since the issue is traceable to the OpenAPI doc?

[zengin]

The issue is mitigated, but it will come back with tomorrow's scheduled generation run, unless I fix microsoftgraph/MSGraph-SDK-Code-Generator#782 beforehand to block releasing the OpenAPI doc.

Even then we need another issue where we fix the root cause parsing error. I don't know where that belongs at the moment as I am not sure if it is the document itself or the parsing library (cc: @andrueastman @irvinesunday @baywet). Only after root causing that, we should close this issue in favor of the underlying issue IMO.

[andrueastman]

This looks to be an issue with the conversion library which looks to be generating incorrect openApi descriptions. microsoft/OpenAPI.NET.OData#243 has been created to follow up on this.

[baywet]

Hey everyone,
Thanks for jumping on the issue, I've been working to root cause that, and discovered a number of issues. Here is a recap of what I've found

1. The OpenAPI.net validation rule is wrong and should look into anyOf/allOf/oneOf before triggering
2. The conversion process is missing the odata type property, and should mark it as required as mandated by the spec
3. The discriminator mapping should list additional ancestors
4. The CSharp code writer don't disambiguate enough when namespaces conflict with type names
5. kiota core generation engine duplicates some properties (this might be a side effect of the incorrect metadata
6. kiota java weekly generation has been failing due to a fault in one of the generation scripts

Now, before Mustafa added the validation step (which I believe should be using hidi instead), the actual blocking step was the disambiguation one in CSharp, which I think we should prioritize. Kiota does log errors on invalid OpenAPI documents but doesn't fail. Of course we should also prioritize fixing the OpenAPI.net and the conversion lib (1, 2, 3) as it impacts more than just Kiota at this point and other tools/apps might choose to completely fail on invalid descriptions (devx API).

I hope this helps everyone see the bigger picture, don't hesitate if you have questions and/or comments!

[andrueastman]

I think we can close this one for now as microsoft/OpenAPI.NET.OData#243 has been resolved and the errors are no longer reported on parsing the openApi document.

cc @baywet

[baywet]

Closing. Can you help make sure a release of hidi with the updated conversion library is out today so the generation pipeline tomorrow takes advantage of all the fixes please?

[andrueastman]

Closing. Can you help make sure a release of hidi with the updated conversion library is out today so the generation pipeline tomorrow takes advantage of all the fixes please?

No worries. Will do