id(world_net): the lib-test target does not compile on main — info_details_rejects_malformed_active_module_hash calls a nonexistent inject_json helper (14 errors)

Summary

On current origin/main, cargo test -p id --lib fails to compile. The test added with the #6633 fix, info_details_rejects_malformed_active_module_hash (pkgs/id/src/world_net.rs:3426-3519 at 4e92da9311), drives the client through a helper that does not exist:

• 7 × error[E0599]: no method named 'inject_json' found for struct 'world_net::WorldClient' — client.inject_json(reply).await at (snapshot lines) :3443/:3455/:3466/:3478/:3490/:3502/:3514. There is no inject_json anywhere in pkgs/id (grep -rn "fn inject_json" = 0 hits on origin/main).
• 7 × error: cannot find macro 'json' in this scope — the test uses bare json!({...}) without importing serde_json::json in the test module.

Total: error: could not compile 'id' (lib test) due to 14 previous errors. The production half of the same change (the info_details match-and-bail on a mistyped active_module_hash) is present and correct — only its regression test is broken. The breakage arrived in the repo-keeper snapshot 4ecb3510c9 ("snapshot code 2026-11-06 17:23 CST"), which swept the in-progress fix into main.

Evidence (verified this run)

• Laptop, worktree at 4ecb3510c9 (== origin/main at the time): cargo test -p id --lib --no-run → the 14 errors above, every one inside this single test; no other file errors.
• Re-verified at fresh origin/main 4e92da9311 (only intervening commit is the unrelated PR org-run: setup job on Blacksmith by default #6639 merge): the test is still present at world_net.rs:3426, and fn inject_json still has 0 definitions in the crate.
• Control: with that one test removed in the probe worktree (locally only, never committed), the lib-test target compiles and runs — 2,313 tests selectable, and the run's unrelated probe test passed. The breakage is exactly this test, nothing else.

Why it matters

While the lib-test target does not compile, no id unit test can run — locally, in a probe worktree, or in CI. Every verification of the pkgs/id work that dominates this repo's current queue (the world_net reply-validation family included) is blocked or forced into workarounds, and the #6633 fix itself has no working regression coverage: the test that was supposed to pin it cannot execute.

Acceptance criteria

• cargo test -p id --lib compiles on main.
• The id(world_net): info_details reads a mistyped 'active_module_hash' as Ok(None) — a malformed module_info reply reports "no module" for a world that has one #6633 coverage intent is preserved by a working test: info_details against scripted replies — string hash → Some, null → None, missing → None, number/boolean/array/object → Err "malformed active_module_hash". Drive it the way the neighboring world_net tests do (a scripted/rogue host via Router + ProtocolHandler speaking the world ALPN, or a real #[cfg(test)] frame-injection helper on WorldClient if that is the intended design — implemented, not just called).
• The malformed-hash cases from the broken test (123, true, ["abc"], {"hash": "abc"}) all appear in the working test.
• No production-code change is needed: the info_details fix itself stands.

Dedup

gh issue list --search (all states) on 2026-11-06 for "inject_json" and for world_net test compile breakage: no hits; gh pr list --search 6633: no PR. Distinct from #6633 itself (the reply-validation bug, whose production fix has landed on main via snapshot — this issue is only its broken test). Related: #6641 (filed this run) hit this breakage as a probe blocker and worked around it locally.