v6: groups/UserType spells db_env_user as db_end_user, and is an unreferenced duplicate

从 io/weaviate/client6/v1/api/rbac/groups/UserType.java 开始，并将其与 rbac/users/UserType 进行比较。使用针对 src/ 的已报告 grep 来验证是否引用了 groups enum，然后确认所选的修正或删除与 public API 及服务器值 db_env_user 一致。重复项和 wire-value 的行为得到一致解决后即表示完成。

Summary

rbac/groups/UserType spells one of its wire values db_end_user. The server's value is db_env_user — env, not end. So DB_ENV_USER can never be produced by a read, and writing it would emit a value the server rejects.

It is also unreferenced, which is almost certainly why the typo survived: nothing in src/main, src/test or src/it uses this enum. That makes it a latent defect rather than a live one, and it may be that the right fix is deletion rather than a correction — see below.

Where it comes from

io/weaviate/client6/v1/api/rbac/groups/UserType.java (6.3.1) — the whole file:

public enum UserType {
  @SerializedName("db_user")
  DB_USER,
  @SerializedName("db_end_user")   // <-- server value is db_env_user
  DB_ENV_USER,
  @SerializedName("oidc")
  OIDC
}

The server's enumeration, entities/models/d_b_user_info.go:142:

if err := json.Unmarshal([]byte(`["db_user","db_env_user"]`), &res); err != nil {

The constant is named DB_ENV_USER, so the intent was clearly db_env_user; only the string is wrong.

It is a duplicate, not the live one

There are two public UserType enums in sibling packages:

grep -rn "UserType" src/ returns the users copy everywhere and the groups copy nowhere. (GroupType, in the same package, is used — this is not a case of the whole package being dead.)

Suggested fix

Two options, and I do not want to presume which you want:

1. Delete rbac/groups/UserType. It is an unreferenced duplicate of a type that already exists and already handles this value correctly. The typo is evidence that nothing has ever exercised it. This removes a trap for whoever wires up the groups client next — they would reach for the enum in their own package and get the broken one. It is a public type, so removing it is a breaking change on paper, even though nothing inside the client can be broken by it.

2. Correct the string to db_env_user. Safe and non-breaking, but leaves two divergent copies of the same concept in the tree, which is how this happened.

The attached PR does (2), as the lower-risk default. Say the word and I will switch it to (1).

Version

• java-client 6.3.1
• Weaviate 1.39.0