Open
Conversation
Scaffolding for the shift away from chunk-copy island deletion. No behavior
change yet — reset and admin delete still go through the old pipeline.
- Settings: add island.deletion.housekeeping.{enabled,interval-days,
region-age-days} (defaults off/30/60). Deprecate keep-previous-island-on-reset
and slow-deletion config entries (unbound from config; getters/setters kept
as @deprecated(forRemoval=true) for binary compat until Phase 4).
- PurgeRegionsService: extract scan/filter/delete/player-cleanup logic out
of AdminPurgeRegionsCommand so the command and the scheduler share one
code path. Handles both pre-26.1 (DIM-1/DIM1 subfolders) and 26.1.1+
(sibling world folders) dimension layouts.
- AdminPurgeRegionsCommand: reduced to ~180 LOC, delegates to the service
and retains only the two-step confirmation UX + per-island display.
- HousekeepingManager: new manager wired in BentoBox.onEnable(). Hourly
wall-clock check; runs the purge service across every gamemode overworld
if enabled and interval has elapsed. Last-run timestamp persisted to
<plugin-data-folder>/database/housekeeping.yml regardless of DB backend,
so the schedule survives restarts. Progress logged to console.
- AdminPurgeRegionsCommandTest: stub plugin.getPurgeRegionsService() with
a real service over the mocked plugin so the extraction is exercised
exactly as the command runs.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Running /bbox purge regions confirm on Paper 26.1.1 tripped AsyncCatcher
because PurgeRegionsService.delete() was saving worlds from the async
worker thread, and World.save() is main-thread-only:
IllegalStateException: Asynchronous world save!
at PurgeRegionsService.delete(PurgeRegionsService.java:151)
The pre-refactor command ran the save on the main thread inside execute()
but I collapsed it into the service. Move the save back out of the
service so all callers are responsible for flushing on the main thread
before dispatching the async delete.
- PurgeRegionsService.delete(): no longer calls Bukkit.getWorlds().save().
Javadoc updated to state the caller contract.
- AdminPurgeRegionsCommand.deleteEverything(): call Bukkit.getWorlds()
.forEach(World::save) before scheduling the async delete. Runs on the
main thread since execute() is invoked there.
- HousekeepingManager.executeCycle(): the existing runTask() save was
fire-and-forget — the async cycle could start scanning/deleting before
the save finished. Block via CompletableFuture.join() until the
main-thread save completes.
- AdminPurgeRegionsCommandTest: add regression asserting the service
never calls Bukkit.getWorlds() itself (would have caught this bug).
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Paper rate-limits its built-in "plugin-induced save detected" warning, so after the scan save fired once, the confirm-path save was silent and looked like it wasn't running. Add explicit plugin.log lines on both sides of every World.save() call in the purge code paths (scan, confirm, housekeeping) so operators always see when the save is happening. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Adds /bbox admin purge age-regions <days> to rewrite per-chunk timestamp tables in .mca files so regions become purgable without waiting wall-clock time. The purge scanner reads timestamps from the region header, not file mtime, so `touch` cannot fake ageing. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…urge-regions-reset
IslandsManager.deleteIsland() used to branch on keepPreviousIslandOnReset: false -> evict from cache, enqueue IslandChunkDeletionManager, MultiLib notify, delete DB row. true -> save with deletable=true and fire the deletion event. With the new region-file purge flow (Phase 1), physical cleanup no longer happens inline at all - old islands are left in place with deletable=true and reaped later by PurgeRegionsService / HousekeepingManager. So the hard-path branch goes away entirely: every call with removeBlocks=true now soft-deletes. Consequences in this commit: - AdminDeleteCommand also soft-deletes until Phase 3 splits it on GameModeAddon.isUsesNewChunkGeneration() (new-gen -> soft-delete, void gamemodes -> ChunkGenerator regen). - Nether/End cascade is a no-op in the soft path (nothing touches chunks); PurgeRegionsService.scan already gates nether/end on isNetherIslands/isEndIslands so vanilla-owned dimensions are skipped when the regions are eventually reaped. - keepPreviousIslandOnReset setter/getter remain as deprecated shims (no longer consulted at runtime); Phase 4 removes the field. - The bentobox-deleteIsland MultiLib subscriber is now unreachable from this server's publishers but stays until Phase 4 deletes the deletion infrastructure wholesale. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Phase 2 made reset leave orphaned islands in place with deletable=true until the region purge reaps them. That meant admins walking around a server had no way to tell an orphan from a normal unowned island — /bbox admin info just showed "Unowned" and entering the area was silent. Two visible cues now: - IslandInfo.showAdminInfo() prints a new "deletable: flagged for deletion and awaiting region purge" line when island.isDeletable() is true, right after the purge-protected line. - LockAndBanListener notifies ops (once per entry, same pattern as the existing lock notification) when they step onto an island flagged deletable. Non-ops still see nothing; this is strictly an admin heads-up. The notification state is cleared when the op leaves the island, so walking back in re-triggers it. New locale keys commands.admin.info.deletable and protection.deletable-island-admin in en-US.yml. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
/bbox admin delete used to always call deleteIsland(island, true, uuid),
which after Phase 2 soft-deletes unconditionally. That is the right
behavior for new-chunk-generation gamemodes like Boxed where chunks
are expensive and the region-file purge reaps them later on the
HousekeepingManager schedule. For void/simple-generator gamemodes it
is the wrong behavior — chunks are cheap, admins expect "delete" to
actually delete, and soft-deleted rows would linger forever because
the repainted region files always look fresh to the purge scan.
Branch on GameModeAddon.isUsesNewChunkGeneration():
- true (new-gen): soft-delete via IslandsManager.deleteIsland(),
same as /is reset. Physical cleanup happens later via
PurgeRegionsService / HousekeepingManager.
- false (void/simple): kick off DeleteIslandChunks (which
routes to WorldRegenerator.regenerateSimple with correct
nether/end cascade gating) to repaint the chunks via the
addon's own ChunkGenerator, then hard-delete the island row
immediately. DeleteIslandChunks snapshots the bounds in its
constructor so the row can be removed before the async regen
completes.
Adds IslandsManager.hardDeleteIsland(island): fires the pre-delete
event, kicks members, nulls owner, evicts from cache, deletes the DB
row. Does not touch world chunks — caller handles physical cleanup.
Phase 4 will remove DeleteIslandChunks, IslandDeletion, and the
CopyWorldRegenerator.regenerateCopy seed-world path; regenerateSimple
and the split here survive.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add a second purge mode that reaps region files for any island already flagged as deletable, regardless of region-file age. Exposed as /bbox admin purge deleted and run from HousekeepingManager on a configurable hourly cadence (default 24h) alongside the existing monthly age sweep. Closes the post-reset gap where orphan island regions sat on disk for 60+ days waiting for the age threshold. Fix: evict in-memory chunks via World.unloadChunk(cx, cz, false) on the main thread before the async file delete, otherwise Paper's autosave re-flushes the deleted region files with the stale chunks. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…urge-regions-reset
Paper's internal chunk cache keeps serving stale block data even after the .mca region files are deleted from disk. The chunks only clear on server restart when Paper discards its cache. Deleting the island DB row immediately left a window where players see old blocks but BentoBox reports no island at that location. The deleted sweep (days==0) now adds island IDs to a pendingDeletions set instead of removing them from the DB inline. On plugin shutdown (BentoBox.onDisable), flushPendingDeletions() processes the set. If the server crashes before a clean shutdown, the islands stay deletable=true and the next purge cycle retries safely. The age-based sweep (days>0) keeps immediate DB removal with the existing residual-region completeness check, since old regions won't be in Paper's memory cache. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Strip the diagnostic logging added during development that printed file size, removed status, and existsAfter for every .mca deletion. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
This setting was made obsolete by Phase 2 which changed /is reset to always soft-delete. The only remaining references were in AdminPurgeCommand for conditional logging — now simplified to always use tier-based progress reporting. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…vice.java Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
…/AdminPurgeDeletedCommand.java Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
…DeleteCommand.java Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
- Rename `deleteableRegions` → `deletableRegions` in PurgeScanResult and all callers (fixes API surface typo) - Distinguish "none-found" from "purge failed" in AdminPurgeRegionsCommand.deleteEverything() with dedicated locale key - Replace CompletableFuture.runAsync() with Bukkit scheduler in AdminPurgeAgeRegionsCommand (ties task to plugin lifecycle) - Clear notifiedPlayers and deletableNotified on PlayerQuitEvent in LockAndBanListener (prevents unbounded set growth) - Add `commands.admin.purge.failed` locale key in en-US.yml Agent-Logs-Url: https://github.com/BentoBoxWorld/BentoBox/sessions/99688668-0dd8-455e-9002-3f229fbefbdc Co-authored-by: tastybento <4407265+tastybento@users.noreply.github.com>
The minimumRank validation was clamping SETTING flags that use -1 (disabled) as their defaultRank. Since -1 is a valid disabled state for SETTING/WORLD_SETTING flags (Island.isAllowed checks >= 0), restrict the validation to PROTECTION flags only. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…bundle icon The "icon" field in blueprint bundle JSON files now accepts three formats: - Plain material name: "DIAMOND" - Vanilla namespaced material: "minecraft:diamond" - Custom item model key: "myserver:island_tropical" Material.matchMaterial() is tried first, handling plain names and vanilla namespaced keys. If the string contains a colon and is not a recognised vanilla material, it is treated as a custom item model key applied to a PAPER base item via ItemMeta.setItemModel(). getIcon() (returning Material) is preserved for binary compatibility. getIconItemStack() is the new method that returns the full ItemStack with model data applied. Panels now call getIconItemStack() to render icons. Closes #2940 Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
When an admin clicks an item from their inventory to set a blueprint bundle icon, IconChanger now checks ItemMeta.hasItemModel() first. If the item carries a custom model key (e.g. paper[item_model="myserver:island_tropical"]), that NamespacedKey string is stored instead of the plain Material name, preserving the full item model through to getIconItemStack(). Plain material items continue to work exactly as before. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Applies the same logic as BlueprintBundle to Blueprint.java: - icon field changed from Material to String internally - getIcon() preserved (returns Material, backward compatible) - getIconItemStack() added — resolves plain names, minecraft: keys, and custom item-model keys (PAPER base + ItemMeta.setItemModel) - setIcon(Material) and setIcon(String) both fluently return Blueprint BlueprintManagementPanel now calls blueprint.getIconItemStack() so custom-model icons are visible to admins in the management GUI. IconChanger also applies item-model detection in the blueprint branch, so admins can click a datapacked item to set a blueprint icon the same way as a bundle icon. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Agent-Logs-Url: https://github.com/BentoBoxWorld/BentoBox/sessions/4b1a4ffa-5cf3-4123-9b9c-80fb712c76fc Co-authored-by: tastybento <4407265+tastybento@users.noreply.github.com>
…ation SonarCloud flagged 8.9% code duplication and repeated "PAPER" literals. - Add ItemParser.parseIconMaterial() and parseIconItemStack() as the single source of truth for icon string resolution - Both BlueprintBundle and Blueprint now delegate to these methods instead of duplicating the logic - Replace repeated "PAPER" string literals with a DEFAULT_ICON constant in each class - Remove unused Sound import from IconChangerTest Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
…item-model-icon feat: support item model keys and namespaced materials for blueprint bundle icon
…nms-directory chore: remove accidentally committed `.paper-nms/` directory and add to `.gitignore`
…lags, cleanup - Version 4.0.0 → 3.15.0; all @SInCE 3.14.0 → @SInCE 3.15.0 - hardDeleteIsland: fire Reason.DELETED post-event so hooks (Level, BlueMap) are notified consistently with the soft-delete path - HousekeepingManager.saveAllWorlds: replace unbounded join() with a 2-minute timeout to avoid hanging the async cycle thread indefinitely - Split single housekeepingEnabled flag into two independent flags: deleted-sweep.enabled (default true — safe, only reaps reset orphans) and age-sweep.enabled (default false — opt-in, more aggressive). Admins get automatic cleanup of reset islands with no config needed. - AdminPurgeDeletedCommand: remove spurious implements Listener / registerListener (class has no @eventhandler methods) - LockAndBanListener.notifyIfDeletable: lambda → Island::isDeletable method ref - Settings: restore keepPreviousIslandOnReset as @deprecated(forRemoval=true) no-ops returning false to preserve binary compatibility for external addons Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
… javadoc The DELETED event already fires immediately in deleteIsland() when the island is marked deletable — addons (OneBlock, Level, etc.) receive it at /is reset time, not when the housekeeping eventually reclaims the region files. - Fix DELETED javadoc: clarify it fires at logical-delete time (soft-delete or hard-delete), not "after all island chunks deleted" as the old wording said - Add Reason.PURGED: fires when the region files AND database row are both physically gone. Fires from: - PurgeRegionsService.flushPendingDeletions() (deleted-sweep, at shutdown) - PurgeRegionsService.delete() age-sweep branch when the DB row is removed - For age-swept islands that were never soft-deleted (isDeletable == false), also fire DELETED before the DB row is removed so addons can clean up data for islands that were pruned by age rather than by an explicit reset Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
…gions-reset Purge region files for soft-deleted islands
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
# Conflicts: # build.gradle.kts
Agent-Logs-Url: https://github.com/BentoBoxWorld/BentoBox/sessions/19b02680-6fb9-456c-9cae-a1c06b2d716b Co-authored-by: tastybento <4407265+tastybento@users.noreply.github.com>
…y /purge unowned The old /purge command only soft-flagged islands as deletable and left the .mca region files in place. Disk wasn't freed until the next housekeeping sweep (or a manual /purge regions). This merges the region-file deletion directly into /purge <days> so admins get the full reap in one step, and removes now-redundant subcommands. Changes: - /purge <days> now delegates to PurgeRegionsService: scans for region files older than N days whose islands aren't protected/spawn/unowned, lists each island location to the log, and reaps the .mca files on confirm. - Dropped /purge regions (merged in), /purge status, /purge stop (regions approach is synchronous from the admin's POV — no state to supervise). - /purge unowned: simplified to flag orphan islands deletable via IslandsManager.deleteIsland(island, true, null); logs each orphan's location; tells the admin to run /purge deleted to reap. - /purge deleted unchanged — still reaps anything already flagged. - Locale cleanup: removed keys for the dropped subcommands; updated description to reflect disk-freeing behavior; added unowned.flagged. - Tests: AdminPurgeCommandTest rewritten end-to-end against a real PurgeRegionsService with a tempDir region layout; AdminPurgeRegionsCommandTest removed; AdminPurgeUnownedCommandTest gains coverage for already-deletable islands. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
- Drop always-true `this.user.equals(user)` guards in AdminPurgeCommand and AdminPurgeDeletedCommand — `this.user = user` was assigned on the prior line, so the check could never fail. - Parameterize the three days-validation tests (notanumber / 0 / -3) as a single @ParameterizedTest. - Extract `wireIsland(id, deletable, purgeProtected, spawn)`, `wireEmptyGrid()`, and `createRegionFile()` helpers in AdminPurgeCommandTest to collapse ~15-line setup blocks duplicated across the 8 scenario tests. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
…ommands AdminPurgeCommand and AdminPurgeDeletedCommand each carried their own inPurge/toBeConfirmed/lastScan state machine and near-identical canExecute / scan-and-prompt / deleteEverything bodies (~75 lines apiece, diverging only in scan source, log prefix, confirm message and an optional chunk-evict step). Pull the shared scaffolding into a package-private AbstractPurgeCommand, with subclasses supplying: - logPrefix() — log line prefix - successMessageKey() — locale key sent on a successful delete - sendConfirmPrompt() — main-thread prompt(s) before user types confirm - beforeDelete(scan) — optional pre-delete hook (used by /purge deleted to evict in-memory chunks before the async file delete) - logScanContents(islands, scan) — optional scan-time logging Behaviour is preserved: same locale keys, same log prefixes, same async threading. The "after a non-empty scan" tail on the failed-delete log line in /purge deleted is dropped — it was always true (we'd already returned on empty) and made the message harder to share. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Merge /purge regions into /purge; drop status/stop; simplify /purge unowned
|
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
No description provided.