Originally, the permissions checking was implemented with a dependency on various permissions plugins, but since the introduction of SuperPerms, it hasn't been needed, and this residue is safe to remove.
Removes the dynamic registration of permissions for classes and arenas.
This means that it is no longer enough to just check whether a Player.hasPermission("mobarena.classes.Knight"), but we also need to check if the Player.isPermissionSet("mobarena.classes.Knight"), which it won't be by default.
To make the stringly typed parts of this dynamic permissions checking more focused, ArenaClass and Arena now both have hasPermission(Player) methods, so the actual permissions checking happens on the objects themselves, rather than from the caller of these methods.
The /back command in Essentials - and similar commands from other plugins that allow you to warp back to your location of death - makes it possible to re-enter the arena after dying.
This commit builds upon the "leaving players" of commit 02f75d0e7f and introduces a more general notion of players being "moved". The arena is "moving" a player if the player is being teleported somewhere by MobArena (to any arena warp), and in this case, all teleports of a player are permitted. This takes care of the cases where MobArena "might be doing it". The rest of the cases are due to commands, plugins, ender pearls, etc., and they are a little easier to deal with separately. The new logic is more straightforward and should prevent most exploits and permit most legit warp cases.
This (hopefully) fixes#313.
Swaps out the hard dependecy on ItemStack for class armor, such that the ThingManager - and thus every ThingParser registered within - has a chance at providing "armor" for a class.
To distinguish armor from items, ArenaMasterImpl prepends "armor:" or slot-specific prefixes to these values before passing them to the ThingManager. This makes it possible to distinguish between items and armor on the thing parser end if applicable - it doesn't make sense for money and command parsers, but they will just end up skipping these values.
The ItemStackThingParser is given a bit of an overhaul. It now depends internally on ItemStack parsers, which can be registered via the ThingManager. This allows custom plugins to hook into the ItemStack-specific part of the ItemStackThingParser to avoid having to duplicate code. Plugins that just want to provide ItemStacks can use this parser approach, and plugins that want to provide more abstract Things can use the Thing parser approach.
Swaps out the hard dependecy on ItemStack for class items (not armor), such that the ThingManager - and thus every ThingParser registered within - has a chance at providing "items" for a class.
Armor is not handled in this commit, because it appears to be quite a bit of extra work, and the "equippable" nature of armor needs to be handled differently than regular inventory items.
The getLogo() method on ArenaClass is removed, because it is no longer used (internally at least). It is residue from the old Spout support.
Turns out that a player who disconnects while in the arena is actually still online if you ask the Player object. The setScoreboard() javadoc specifically says it throws an IllegalStateException in that case, so we can just catch and swallow it.
It's unclear what the state of the player's scoreboard is when they rejoin the server, but a borken scoreboard is better than a rogue exception.
This commit removes certain items from the in-arena inventories of players who choose the "My Items" class:
- Ender chests allow players to smuggle stuff out of arenas
- Ender pearls can be used to warp to unintended spots, and to build ender chests
- Shulker boxes can carry any items - including ender chests
- Shulker shells can be used to build shulker boxes
At the end of the day, ender chests are the bad seed. The rest are just to prevent ender chests from making appearances.
Apparently, the /give command will drop an item "as" the target player, resulting in the player picking up the item if there is room in their inventory. This triggers a PlayerDropItemEvent, which MobArena catches and tries to cancel if the target player is in an arena. Furthermore, if the player is a spectator outside of the arena region, MobArena will force the player to leave the arena. This triggers the reward granting logic, so using the /give command as a reward has the potential of causing an infinite loop.
This commit introduces the idea of "leaving players", i.e. players in the process of leaving the arena. In the event of a dropped item from a player who is currently leaving an arena, the PlayerDropItemEvent is ignored, because it is assumed to be from the /give command. Note that this doesn't actually prevent normal PlayerDropItemEvents from causing a forced leave, since the player won't be in the process of leaving at that specific point.
This commit changes the naive behavior of the ConfigUtils methods addIfEmpty() and addMissingRemoveObsolete(). Instead of an almost guaranteed config-file write on every invocation of either method, writes only happen if the loaded configuration changes. The excessive writes result in long config reload times, and this change fixes that, effectively fixing the second part of #435.
To further the performance boost, the resource reads are cached in a map - this turns out to not be an issue for server plugin reloads, as a new ClassLoader instance is used to load the new set of plugins.
This means that the cleanup code will only run if there are players in the lobby, arena, or spectator area. This fixes the first part of #435 where the cleanup code is taking too long for large/many arenas.
Originally, the force end command was meant as a way to circumvent any condition keeping players from leaving the arena and cleaning it up. In retrospect, the main reason for using force end is to "get people out of there", but since there's plenty of stuff that can go wrong when a player leaves, this isn't really that helpful, as exceptions will just cause the command to break at the same point anyway.
Use ItemStack display name in ItemStackThing (if available).
This means that named items will appear in the rewards list with their name rather than the ItemStack type. This is useful for "tokens" type items.
Note that named items are still not supported by the built-in item parser, so this commit only affects ItemStackThings created by custom parsers (for now).
This commit updates the Arena Setup documentation.
* Follows RST document style guide
* Uses active voice / minor edits
* Improve document organization
I also renamed the wave-formulas.rst file since it had a typo, and was
referenced in the Arena Setup document.
This commit does the following:
* Uses ReStructuredText style guide for the announcements documentation
* Uses active voice / minor edits
* Adds examples
This was tested locally in a development environment.
With CommandThing and its parser, it is now possible to use commands anywhere a thing can be used. Commands are invoked as the console/server, and they support a single variable, the name of the recipient player.
Commands are give-only, meaning they will fail to be "taken" from players, and they cannot be "held" either. The idea of commands as things basically only makes sense in the context of rewards.
While it doesn't really make sense to have class prices be a Thing (or maybe it does?), this does get rid of the dirty "ItemStacks wit ID -29" hack for economy money, which will eventually break, when the upstream int-based ID API breaks.
With this commit, the Things API is introduced to the code base, but it is not yet used. It introduces the building blocks for an extensible architecture that supports custom parsers and custom "things". This should allow other developers simple, yet powerful hooks into the way MobArena handles class "equipment" and rewards. The basic skeleton includes parsers for ItemStacks and economy money, so it should be interchangeable with the current inner workings of the plugin.
The commit also adds an overload to the ItemParser that allows for a the method to fail silently. This is necessary to avoid false negatives in the log in case the ItemStackThingParser fails but a different parser succeeds.
The purpose of this is to actually commit to version control the jar(s) necessary to compile. It's not a best practice whatsoever, but if we can commit gradle wrappers, we can commit hard-to-incorporate binary dependencies.
As a nice side effect, it should be possible to run some basic CI tooling on the repository.
