5113. Validation of oneof nested objects

[optician]

References to other Issues or PRs

Fixes #5113

Have you read the Contributing Guidelines?

Yes.

Brief description of what is fixed or changed

The fix follows the suggestion from the issue. The oneOf check is moved before msgValue.Mutable() resets other option of this oneOf.
Also, previous behavior didn't validate nested messages correctly. After the cycle "fieldDescriptor" points to a leaf of a path tree and it's not a member of a oneOf field. For example, the piece of a proto

 ...
  oneof nested_oneof_value {
    int32 nested_oneof_int32_value = 50;
    Proto3Message nested_oneof_value_one = 46;
  }
 ...

and the path ["nested_oneof_value_one", "int32Value"]. WhichOneof(nested_oneof_value_one) returns nested_oneof_value, but `WhichOneof(int32Value) returns nil.

Other comments

I struggled with codegen refreshing and spent quite a lot of time on a simple task. The contribution instruction does not regenerate *.pb.go files when protos are changed (and runs very slowly because it recompiles protoc). Other documentation also lacks instructions. I hope I found the correct way and it will save time for others.

[johanbrandhorst]

Minus the contribution guide changes, this looks great, thank you so much!

[johanbrandhorst]

Seems like some swagger files still need to be regenerated, could you try running it again? See the failures in the generate CI job.

[optician]

As I understand the file shouldn't change. It seems ok now. I can't explain what happened in the previous iteration 🤷

[johanbrandhorst]

Thanks for your contribution!