fix(ShootingStar): close stop race on null panel and null state (1.4.11)

[chsami]

Problem

Client error telemetry for ShootingStarPlugin 1.4.10 (client 2.6.33) recorded 793 NullPointerExceptions across three groups. All three come from stopping the plugin:

• 2529 (787 events): NullPointerException on the AWT EDT in ShootingStarPlugin.lambda$updatePanelList$3:419, where panel.updateList(stars) runs with panel already null.
• 2532 (1 event): the same NPE in lambda$updatePanelList$4:423, from panel.refreshList(stars).
• 2531 (5 events): NullPointerException in ShootingStarScript.lambda$run$1:128 on switch (state), after state had been set to null.

Root cause

1. Panel NPE. When the player died and shutdownOnDeath was on, the death handler called shutDown() directly. That removed the panel and set panel to null, but the plugin was still registered with the event bus. Later events and game ticks kept calling updatePanelList, and each call queued an invokeLater lambda that dereferenced the null panel on the EDT. A runnable queued just before a normal stop could hit the same NPE.
2. State NPE. Script.shutdown() calls mainScheduledFuture.cancel(true), and that does not wait for an iteration that is already running. ShootingStarScript.shutdown() then set state, currentStar and pickaxe to null, so the iteration still running read null fields at switch (state). The same could happen with currentStar in WALKING, MINING and BANKING.

Change

• The death handler now calls Microbot.stopPlugin(this), so the plugin is stopped and unregistered the normal way. Called off the EDT, this does not block.
• updatePanelList returns early when panel is null. The EDT lambda reads panel into a local once and skips the update if it is null.
• shutdown() no longer clears state, currentStar or pickaxe. run() resets them at startup instead, next to the other per-run fields it already resets.
• The overlay reads the static state into a local once before it null-checks and renders it.
• Rs2Antiban.getCategory().isBusy() goes through a null-safe isAntibanBusy helper, because Rs2Antiban.resetAntibanSettings() sets the category to null.
• Version bumped from 1.4.10 to 1.4.11.

Validation

• Added ShootingStarStopRaceTest (5 tests). It covers a queued panel update after the panel is removed, which NPEs on the old lambdas. It also checks that shutdown() leaves the loop state alone for an iteration that is still running, which fails on code that clears the state on stop.
• ./gradlew test --tests 'net.runelite.client.plugins.microbot.shootingstar.*' build -PpluginList=ShootingStarPlugin passes. That covers ShootingStarStopRaceTest (5), ShootingStarWalkBeforeHopTest (15) and ShootingStarWorldFilterTest (10).

Known gaps

• The script executor is reused across restarts. If the plugin is stopped and then restarted straight away while an old iteration is still running, run() resetting state could still cause an NPE in that old iteration. The main loop catches and logs it, and it does no harm.
• Rs2Antiban.resetAntibanSettings() setting the category to null is a client issue. JewelryScript and GlassblowingScript call Rs2Antiban.getCategory().isBusy() the same way and are not covered here. A client-side non-null default category would fix all of them.
• Group 2530 (Rs2Bank.walkToBank NPE on a null bank location) arrived in the same upload. Client fix #1996 already covers it, so this PR does not touch it.

🤖 Generated with Claude Code

https://claude.ai/code/session_01LjtEMzQHGuq4M1Mpqk9pPV