Skip to content

Changelog

Replace the jar and restart. Your config is kept as it is.

  • A developer API, like CoreProtect’s, so other plugins can work with xRewind: they can log the blocks and items they change (so a plot reset shows in lookups and can be undone), look up the history, preview, apply and undo rollbacks and restores, and hold the writer for a moment while a backup copies the database. Plugins’ lookups and rollbacks run like yours: in the background, a few milliseconds per tick, one rollback at a time. See Developer API. It costs nothing unless a plugin uses it.
  • Plugins can listen for rollbacks (and stop one before it starts), and for lookups, to keep an audit log of who looked up what, staff commands included. A staff member whose rollback, apply or undo another plugin stopped is told so.
  • A purge or a CoreProtect import that the server stops partway says so, with how many changes the purge deleted by then.
  • Clicking Next page below a lookup’s results shows the next page. It ran <next> instead, which isn’t a command.
  • After a purge, the space is handed back to the disk a few thousand pages at a time, so logging never waits on all of it at once.
  • A CoreProtect import can’t wait forever for its last rows when one of them is a change the database won’t take.
  • /xrewind reload works on an empty server. Paper pauses a server that’s been empty for a minute, and the reload waited for the next player to join.
  • A rollback or restore typed in the console of an empty server runs straight away, not when the next player joins.
  • On MySQL, a connection whose database stops answering gives up after five minutes, or ten seconds to connect, instead of holding the writer forever.

Replace the jar and restart. Your config is kept as it is.

  • A rollback marks as rolled back only the changes it really made. Before, changes it skipped (a block this Minecraft can’t read, a world that isn’t loaded, or the rest of a rollback that stopped partway) were marked too, so no later rollback could reach them.
  • A database that can’t be reconnected to no longer leaks a connection on every try, which on MySQL could use up its connections.
  • Three write failures hours apart no longer count as “three in a row”: only failures back to back make xRewind write a batch row by row.
  • /xrewind page goes up to page 500. Much deeper pages read thousands of rows to show seven.
  • On MySQL, starting again after a database update that stopped partway carries on, instead of failing on the tables already made.
  • A huge TNT chain no longer slows down as xRewind tidies its list of recent breaks, and a preview a player left without applying is let go of.

Replace the jar and restart. Your config is kept as it is.

  • A rollback that takes under a second says how long to the hundredth: “in 0.04s”, not “in 0.0s”, which read as if nothing happened.
  • /execute as <player> run xrewind ..., and command blocks running as a player, now measure r: from that player and can #preview to them. Before, they were treated as the console.

The first release. Drop the jar in and start the server: it logs to SQLite in its own folder, with nothing to set up. Coming from CoreProtect? Run /xr import coreprotect to bring your history.

  • Blocks placed and broken, with what they held (a chest’s items, a sign’s text), and the torches, signs and rails that fall off with them.
  • Explosions (TNT is put down to whoever lit it), fire, buckets and liquids that destroy something, pistons, and mobs that change blocks (endermen, withers, falling sand).
  • Items put in and taken out of containers, kills, sign text, and joins and quits.
  • Off unless you turn them on: hopper transfers, door and lever clicks, leaf decay, chat and commands.
  • Each kind is a switch under log, and one that’s off isn’t even listened for. excluded-worlds leaves worlds out.
  • Logging costs about 40 nanoseconds per change on the game’s threads: everything else happens on one background thread, in batches.
  • /xr inspect: click a block to see its history, or place one to see what was there.
  • /xr lookup with CoreProtect’s parameters: u: t: r: a: i: e: w: and #count. /xr page and /xr near.
  • /xr rollback and /xr restore with the same parameters, #preview to see the result first, and /xr undo.
  • Blocks come back with what they held, and items taken from containers are put back.
  • Blocks are set a few milliseconds per tick, chunk by chunk (region by region on Folia), so even a big rollback doesn’t freeze the server.
  • A rollback needs t:, covers 10 blocks around you without r:, and past 100 blocks needs xrewind.rollback.global.
  • Automatic purges, free: history older than 30 days is deleted once a day (purge.keep-days; 0 keeps everything), and /xr purge.
  • SQLite by default, MySQL or MariaDB for networks. No driver is bundled: Paper and Folia include them.
  • /xr import coreprotect reads CoreProtect’s database (SQLite or MySQL) and imports blocks, kills, containers, sessions, signs, chat and commands.
  • /xr status shows where the history goes and whether logging is keeping up.