Files
Bubberstation/code/datums/move_manager.dm
Leland Kemble 61fb177584 Fixes several ways you could become permanently immobile (#96112)
## About The Pull Request

When the `force_move()` component is applied, it registers two signals
on the involved mob, to stop it moving, and a third one on the
`move_loop` it's created. The signal on the created loop is there so
that when the loop ends, the `force_move` component can delete itself
and free the user. However, there are no checks for whether the
`move_loop` was successfully created(which isn't the actual problem,
although I have made it runtime there so it is easier to catch in the
future). The issue is that when creating two identical moveloops, it
cancels the creation of the loop, returning nothing even though there is
a loop in effect. The fix was to return the existing loop in this
scenario, because it's identical to what would have been made had there
not been a loop already.

## Why It's Good For The Game

fixes #96069 - Certain raptors transfer hits to their owners, and (this
seems like an oversight in and of itself) piercing projectiles thus hit
twice, which in the case of the honk staff does two slips, two
`force_move`s, two loops, and triggers our identical loop problem.
fixes #88352 - The issue report is vague and lacking in specific
scenarios, but the mention of slipping again fixing it(which it does
with this bug, as the new `force_move` component deletes the broken one)
makes me pretty sure it's the same bug
fixes #95994 - I'm not certain on this one because I don't know how the
Parade works internally, but it seems likely that it could cause two
slips on the same tile
fixes #94515 - I'm just assuming this is the same bug based on it having
the same symptoms

## Changelog
🆑

fix: Several ways you could become permanently immobile, primarily based
around lube & other methods of slipping, have been fixed

/🆑
2026-05-23 17:13:25 +02:00

174 lines
7.1 KiB
Plaintext

/**
* Acts as a namespace for movement packet/type related procs
*
* Exists to provide an in code implementation of movement looping
* Replaces things like walk() or walk_to(), among others
*
* Because we're doing things in engine, we have a lot more control over how different operations are performed
* We also get more say in when things happen, so we can subject movements to the whims of the master controller
* Rather then using a fuck ton of cpu just moving mobs or meteors
*
* The goal is to keep the loops themselves reasonably barebone, and implement more advanced behavior and control via the signals
*
* This may be bypassed in cases where snowflakes are nessesary, or where performance is important. S not a hard and fast thing
*
* Every atom can have a movement packet, which contains information and behavior about currently active loops, and queuing info
* Loops control how movement actually happens. So there's a "move in this direction" loop, a "move randomly" loop
*
* You can find the logic for this control in this file
*
* Specifics of how different loops operate can be found in the movement_types.dm file, alongside the [add to loop][/datum/move_manager/proc/add_to_loop] helper procs that use them
*
**/
/datum/move_manager
GLOBAL_DATUM_INIT(move_manager, /datum/move_manager, new)
///Adds a movable thing to a movement subsystem. Returns TRUE if it all worked, FALSE if it failed somehow
/datum/move_manager/proc/add_to_loop(atom/movable/thing_to_add, datum/controller/subsystem/movement/subsystem = SSmovement, datum/move_loop/loop_type, priority = MOVEMENT_DEFAULT_PRIORITY, flags, datum/extra_info)
var/datum/movement_packet/our_data = thing_to_add.move_packet
if(!our_data)
our_data = new(thing_to_add)
var/list/arguments = args.Copy(2) //Drop the atom, since the movement packet already knows about it
return our_data.add_loop(arglist(arguments))
///Returns the subsystem's loop if we're processing on it, null otherwise
/datum/move_manager/proc/processing_on(atom/movable/packet_owner, datum/controller/subsystem/movement/subsystem)
var/datum/movement_packet/packet = packet_owner.move_packet
if(!packet)
return
var/datum/move_loop/linked_loop = packet.existing_loops[subsystem]
if(!linked_loop)
return
if(linked_loop.flags & MOVEMENT_LOOP_IGNORE_PRIORITY)
return linked_loop
if(linked_loop != packet.running_loop)
return
return linked_loop
///A packet of information that describes the current state of a moving object
/datum/movement_packet
///Our parent atom
var/atom/movable/parent
///The move loop that's currently running, excluding those that ignore priority.
var/datum/move_loop/running_loop
/**
* Flags passed from the move loop before it calls move() and unset right after.
* Allows for properties of a move loop to be easily checked by mechanics outside of it.
* Having this a bitfield rather than a type var means we don't get screwed over
* if the move loop gets deleted mid-move, FYI.
*/
var/processing_move_loop_flags = NONE
///Assoc list of subsystems -> loop datum. Only one datum is allowed per subsystem
var/list/existing_loops = list()
/datum/movement_packet/New(atom/movable/parent)
src.parent = parent
parent.move_packet = src
/datum/movement_packet/Destroy(force)
parent.move_packet = null
parent = null
for(var/datum/controller/subsystem/processor as anything in existing_loops)
var/datum/move_loop/loop = existing_loops[processor]
if(QDELETED(loop))
continue
qdel(loop)
existing_loops.Cut()
existing_loops = null //Catch anyone modifying this post del
return ..()
///Adds a loop to our parent. Returns the created loop if a success, null otherwise
/datum/movement_packet/proc/add_loop(datum/controller/subsystem/movement/subsystem, datum/move_loop/loop_type, priority, flags, datum/extra_info)
var/datum/move_loop/existing_loop = existing_loops[subsystem]
if(existing_loop && existing_loop.priority > priority)
if(!(existing_loop.flags & MOVEMENT_LOOP_IGNORE_PRIORITY) && !(flags & MOVEMENT_LOOP_IGNORE_PRIORITY))
return //Give up
if(existing_loop?.compare_loops(arglist(args.Copy(2))))
return existing_loop //it already exists stop trying to make the same moveloop
var/datum/move_loop/new_loop = new loop_type(src, subsystem, parent, priority, flags, extra_info) //Pass the mob to move and ourselves in via new
var/list/arguments = args.Copy(6) //Just send the args we've not already dealt with
var/worked_out = new_loop.setup(arglist(arguments)) //Here goes the rest
if(!worked_out)
qdel(new_loop)
return
existing_loops[subsystem] = new_loop
if(existing_loop)
qdel(existing_loop) //We need to do this here because otherwise the packet would think it was empty, and self destruct
contest_running_loop(new_loop)
return new_loop
///Attempts to contest the current running move loop. Returns TRUE if the loop is active, FALSE otherwise
/datum/movement_packet/proc/contest_running_loop(datum/move_loop/contestant)
var/datum/controller/subsystem/movement/contesting_subsystem = contestant.controller
if(contestant.flags & MOVEMENT_LOOP_IGNORE_PRIORITY)
contesting_subsystem.add_loop(contestant)
return TRUE
if(!running_loop)
running_loop = contestant
contesting_subsystem.add_loop(running_loop)
return TRUE
if(running_loop.priority > contestant.priority)
return FALSE
var/datum/controller/subsystem/movement/current_subsystem = running_loop.controller
var/current_running_loop = running_loop
running_loop = contestant
current_subsystem.remove_loop(current_running_loop)
if(running_loop != contestant) // A signal registrant could have messed with things
return FALSE
contesting_subsystem.add_loop(contestant)
return TRUE
///Tries to figure out the current favorite loop to run. More complex then just deciding between two different loops, assumes no running loop currently exists
/datum/movement_packet/proc/decide_on_running_loop()
if(running_loop)
return
if(!length(existing_loops)) //Die
qdel(src)
return
var/datum/move_loop/favorite
for(var/datum/controller/subsystem/movement/owner as anything in existing_loops)
var/datum/move_loop/checking = existing_loops[owner]
if(checking.flags & MOVEMENT_LOOP_IGNORE_PRIORITY)
continue
if(favorite && favorite.priority > checking.priority)
continue
favorite = checking
if(!favorite) //This isn't an error state, since some loops ignore the concept of a running loop
return
var/datum/controller/subsystem/movement/favorite_subsystem = favorite.controller
running_loop = favorite
favorite_subsystem.add_loop(running_loop)
/datum/movement_packet/proc/remove_loop(datum/controller/subsystem/movement/remove_from, datum/move_loop/loop_to_remove)
if(loop_to_remove == running_loop)
running_loop = null
remove_from.remove_loop(loop_to_remove)
if(loop_to_remove.flags & MOVEMENT_LOOP_IGNORE_PRIORITY)
remove_from.remove_loop(loop_to_remove)
if(QDELETED(src))
return
if(existing_loops[remove_from] == loop_to_remove)
existing_loops -= remove_from
decide_on_running_loop()
return
/datum/movement_packet/proc/remove_subsystem(datum/controller/subsystem/movement/remove)
var/datum/move_loop/our_loop = existing_loops[remove]
if(!our_loop)
return FALSE
qdel(our_loop)
return TRUE