The blockList() call returns the actual List<Block> object in the explode event, which means it'll be shared between the the fake event and the original event. As a result, the call to blockList().clear() will clear the shared list, and the following call to blockList().addAll(fake.blockList()) results in trying to add the empty list to itself.
This commit makes sure to copy the original event's block list before sending it to the fake event.
Okay, "Fixes" would be lying, and that's terrible, but here's something to help you shrug it off: The timer module is a bit of a mess, and since there isn't actually anything "dangerous" about this, let's just calm our tits and focus on other stuff.
Before, these redundant calls didn't really do anything. Now they mess everything up, because setContents() doesn't set inventory contents only, but the entire inventory instead, because that's so super useful. Well done, boys!
#notsalty
Because it is clearly impossible to change JavaDocs to correctly describe what a method does, the method setContents() has now been changed to a useless, over-aggressive, inventory-wiping piece of junk that must be invoked before anything else. This commit ensures that armor contents are set after the setContents() method is called, overwriting the nulls that the setContents() method creates.
This changes the format from:
<effect>:<duration>:<amplifier>
to a (pressumably) more user-friendly format of:
<effect>:<amplifier>:<duration>
The reason for this swap is that the duration is more likely to be
wanted left at default than the amplifier. From previous user input
on the matter, it sounds like users generally want the bosses to have
"permanent" or "infinite duration" effects, and this change makes it
a little easier on them, not having to specify a duration.
With this feature you can add potions to the bosses when they spawn,
adding the ability to add strength to the boss for the first 30 seconds
making it harder for the first few seconds and once the effect wears
goes away the boss becomes easier. The syntax is in the same style as
potion items, except it uses either the potion effect enum name or the
potion effect id.
POTION_EFFECT:duration in seconds:amplifer
SLOW:360:4;INVISIBILITY:36000:0
If no duration is given, the plugin will default to sixty seconds. If
the amplifer was not provided, the plugin will default to zero. The
potion effect amplifcation starts at 0 being level 1.
Some of the config-file resets were fixed in a previous commit:
4330885d46
The resets fixed previously were ones where snakeyaml would throw a fit
due to tabs. Because an exception is thrown, the call to saveConfig()
is never reached, which means the config-file is never reset.
However, errors resulting from other illegal characters were not caught
because the built-in reloadConfig() method catches, prints and then
simply does nothing about any exceptions thrown, which means the actual
YamlConfiguration would be empty. Saving this empty file would result
in a reset, because MobArena, during initialization, sees missing bits
and pieces and tries to insert them. Thus, the resets were not an issue
with MobArena's attempt to keep the config-file tidy, but the reckless
handling of errors in the built-in config methods in JavaPlugin.
This commit completely overrides the methods in JavaPlugin that would
result in these resets, and the MobArena plugin class stores its own
reference to the config file and an associated FileConfiguration. When
the reloadConfig() and saveConfig() methods are called, the checked
exceptions are caught and rethrown as unchecked exceptions, causing the
plugin to crash when it needs to do so.
This commit creates a fail-fast exception-throwing guard inside the
method for handling config-reloading in the main plugin class.
If a config-file contains a tab, an exception will be thrown when
the plugin loads (as it did before), which means the plugin will not
load correctly. However, upon reloading with the reload command, an
exception is thrown before the current settings are overwritten.
This fixes a long-standing issue with config-files resetting upon
reload with invalid syntax. Verification is needed, as this commit
has only been tested on OS X where tabs and carriage returns are
injected via vim.
Due to the start-delay-timer holding a reference to the
auto-start-timer as its internal callback, the start-delay-timer
should be stopped, when both timers should be stopped.
The StartDelayTimer wraps the AutoStartTimer and provides a means of
"locking" the lobby until a configurable time has passed. During this
time, it is possible for other players to join the lobby and ready up,
but the arena will not start until the StartDelayTimer has run out,
after which the AutoStartTimer is started (if applicable).
New per-arena setting: start-delay-timer: [seconds]
New announcement: arena-start-delay
This framework consists of interfaces and classes for managing various
kinds of timers. The Timer interface models a timer with configurable
tick intervals. The TimerCallback interface provides a means of adding
logic to the following timer events: onStart, onStop, onFinish and
onTick.
Timer implementations include a CountdownTimer (kitchen timer), and a
StopwatchTimer, which should both be self-explanatory.
The purpose of this framework is to add a means of creating a very
generalized approach to timers, in hopes of abstracting away some of
the Bukkit scheduling for easy maintenance and increased portability.
Future work:
- Port auto-start-timer
- Port boss abilities
- Port spawner
- Create 'delay-timer' as per DBO ticket request.
- Create various types of cooldowns, delays, etc.
When the WaveManager is queried for the next wave to be spawned via
the next() method, it now returns a copy of the wave instead of the
wave itself. This is because waves (boss waves in particular) have
state, and this state is transferred to the next occurrence of the
given wave. This caused recurrent boss waves to inconsistently share
state, which resulted in the BossAbilityThread bailing early after
the first occurrence of the boss wave it belonged to. By copying
the initial wave state in the WaveManager, the issue is fixed.
Fixes: http://dev.bukkit.org/bukkit-plugins/mobarena/tickets/1220/
