Joshua Humphries

It turns out that the production for `oneof` in the [documented grammar](https://developers.google.com/protocol-buffers/docs/reference/proto2-spec#oneof_and_oneof_field) is similarly incorrect. It also claims to allow `emptyStatement`, but protoc does not allow it. ```protobuf syntax =...

FYI, this is closely related to #5075

My 2 cents: option 2 still makes the most sense. While some runtime functions may accept questionable input without complaining, this is a validation framework after all, so it seems...

grpcurl now has an alpine version https://github.com/fullstorydev/grpcurl/pull/311 but grpcui not yet

@residual-paranoia, yes, you can load the descriptors generated by protoc: https://pkg.go.dev/github.com/jhump/protoreflect/desc#CreateFileDescriptorsFromSet First you'll unmarshal the file into a `*descriptorpb.FileDescriptorSet` (since the file contents are the binary encoding of this type)....

`ExpandAny` causes it to use the expanded format for `google.protobuf.Any`. So instead of this: ```protobuf // unexpanded, binary-encoded message as value type_url: "type.googleapis.com/google.protobuf.Duration" value: "\b\xf0\x86\x05" ``` We want this: ```protobuf...

> only observable behavior of the new proto printer Well, it can't do this if it doesn't know about the message type. So it's likely observable if you put in...

This breaks the `protoprint` tests. You can try simply re-generating the golden output files by changing [this var](https://github.com/marcoferrer/protoreflect/blob/7c55a390a25f088ae1483b738199a913d1ed023b/desc/protoprint/print_test.go#L25) to `true`, run the test, and then change it back to `false`....

Yeah. This was really why I put the big note on the README about this using API v1. It's probably best to just punt on this until a future major...

> So I actually got it working. Wanted to see what you thought about the implementation. While I admire your perseverance, I think the result is too much. Given that...