docs/bnstoolkit-discussion-points-2026-06-07.md

@@ -1,8 +1,45 @@
 # bnstoolkit M0 — discussion points for testing session
 
-> Built 2026-06-07 (early hours UTC). Shipped via pack v0.34.0.
+> Built 2026-06-07 (early hours UTC). Shipped via pack v0.34.0 then hotfixed via v0.34.1.
 > Repo: http://10.16.5.102:3000/minecraft/bnstoolkit
-> Jar: `bnstoolkit-0.1.0.jar` (15.6 KB)
+> Jar: `bnstoolkit-0.1.1.jar` (15.6 KB)
+
+## ⚠ Server status: needs a manual Start
+
+The first deploy (`0.1.0`) crashed the dedicated server during mod
+construction — see "Bug caught + fixed" below. The fixed jar (`0.1.1`)
+is already staged in `/srv/fast/brass-sigil-server/server/mods/`, but
+the JVM exited cleanly so systemd didn't auto-restart and I deliberately
+didn't try to brute-force the admin panel credentials.
+
+**To bring the server back: click Start in the web admin panel
+([http://glados:8080](http://glados:8080)) when you wake up.** The new
+boot should load `bnstoolkit-0.1.1` cleanly. Verification:
+
+```
+sudo journalctl -u brass-sigil-server.service --since "2 minutes ago" \
+  | grep -E "bnstoolkit|Done \([0-9]+|ModLoadingException"
+```
+
+Look for `[bnstoolkit] common setup complete` AND no `ModLoadingException`.
+
+## Bug caught + fixed (0.1.0 → 0.1.1)
+
+```
+IllegalArgumentException: class uk.sijbers.bnstoolkit.BnsToolkit
+has no @SubscribeEvent methods, but register was called anyway.
+```
+
+I called `NeoForge.EVENT_BUS.register(this)` in the @Mod constructor
+without yet having any `@SubscribeEvent` methods on the class. NeoForge
+21.1 treats that as fatal at construct-mods time. Removed the call;
+will come back together with the first real event handler in M1.
+
+**Why local `./gradlew build` didn't catch it:** the build step only
+compiles + jars. The construct-mods phase only runs when a real
+NeoForge instance loads the mod. Action item for M1: add a
+`gameTestServer` Gradle run config that boots the dedicated server
+profile in CI-style so this class of bug fails locally next time.
 
 ## What this is