This won't cover all effects of internal changes, some blocks might be
interpreted wrongly still, some shapes may have changed in an
incompatible way (e.g. skulls).
This seems to be more appropriate, for the case of multiple checks
triggering, but also for better performance, in case the check doesn't
even support input-specific penalties.
This is a first step in, which doesn't change the default behavior,
however it might break plugins that rely on certain internals.
PenaltyAction allows to do something with probabilities to consider,
including the possibility to select the first applicable penalty or
applying several penalties. There will be player-specific penalties,
which are applied during ViolationData.executeActions always, and there
will be input-specific penalties, e.g. for applying within the event
listener.
Potentially breaking:
* Return value of executeActions is now void for Action + ViolationData.
* Return value of Check.executeActions is ViolationData now.
* CancelAction is now extending PenaltyAction.
* CancelPenalty may cancel, but might not, due to probability.
* IViolationInfo.hasCancel -> deprecated, now returns willCancel().
* IViolationInfo.willCancel is now used, applicable penalties are
estimated on creation of ViolationData.
* Custom actions can no longer be used to cause cancel. Only penalties
can do so now (due to the return type change). CancelAction is still
there to keep a simple action for canceling.
Not yet:
* InputSpecificPenalty support for fight checks and using them in the
default actions.
* Configuration for penalties (currently only a plugin could override
the action factories, later penalties may have a probability to apply,
reference each other, allow first match, apply several at once).
Introduce an interface to indicate if a CheckDataFactory or IRemoveData
instance can do something better than removing all data, in case the
system time ran backwards. An extra method in data manager is used
instead of clearAllData, which will test for implementation of that
interface. Concerns CheckDataFactory instances accessible via CheckType
and IRemoveData instances registered with DataManager (via
NoCheatPlusAPI or directly).
Implementation details and related changes:
* TickTask: Let ActionFrequency handle time running backwards (spikes).
* NetData/Factory: Use HashMapLOW. Selectively clear/adjust.
* MovingData: Keep past move tracking (and some other).
Remove ambigue method, apply a different default margin, up to the next
block in steps of 0.25, if within 0.35 reach.
Default methods use this correction now, so some places might check with
too high a margin now.
Not entirely sure this will still protect from anything, shortish:
* Remove cancelling due to coarse pre-checks.
* Rather correct the end location back onto the end block somehow.
* Retry ray-tracing with the pitch and yaw of past flying packets.
* Let the direction check handle the off-too-far part and let people
blame that one for remaining amounts of false positives.
Missing:
* Should confine by distance to last move, perhaps.
Use check-specific debug methods for convenience.
Add to: Check, CheckListener, BaseAdapter.
Relay to: CheckUtils.
Side effects:
* Remove constructor: CheckListener().
This is half a guess on base of a request. Hooks can now check if there
are any log actions, or cast IViolationInfo to ViolationData and query
if anything would be logged to a certain stream.
Alterations pending (allow query for multiple streams or just get
streams/configs, what to have in the official API in IViolationInfo).
