original pr here: #59789 (Closed because he didn't think it was good enough)
came back to this because i realized that
all machines were area sensitive, meaning they had a list with at least a reference to themselves (assuming they arent in the contents of another movable which most arent) for the purposes of handling power differences when their area changes
pipes are machines
there are ~14k machines and ~6k pipes
i made this problem worse with a recent pr by making it a nested list
so i needed to track what machines needed power, and this pr had work already done that could be used for that purpose. now machines that have use_power == NO_POWER_USE do not have this extra memory overhead for no reason
currently every machine that uses power draws that amount from its area from a dynamic channel via auto_use_power() which is called every SSmachines fire(), then in apc/process() the area's dynamic power draw is reset and the power is used. with static power its not calculated then reset every loop, its just taken from the grid. so now machines handle updating their static power usage from their current area (this doesnt touch power machines that require a wire connection). in order to allow this, use_power, idle_power_usage, and active_power_usage have setters to track state correctly and update the static power usage on the machines current area and handle area sensitivity.
also goes through a lot of heavy abusers of SSmachine processing time and tries to make it faster. makes airalarm/process() into a signal handler for COMSIG_TURF_EXPOSE since air alarms only need to process for changes.
Why It's Good For The Game
SSmachines isnt the heaviest hitter in terms of total cpu and certainly not in terms of overtime, but its not a lightweight. it frequently takes > 50ms to complete a run and seems to be in the top 5 or so of subsystem costs looking at some round profilers
also gets rid of a few thousand lists since every pipe no longer has two useless lists each (and any other machines that dont use power)
Love ya kyler
Co-authored-by: Rohesie <rohesie@gmail.com>
Simple_animals / mobs are the biggest lie in this code-base. They're far from simple and have an extreme god-object problem. Especially when you get to /hostile, where there is so many procs, vars, and what not, that you can't make any interesting additions without snowflaking the hell out of the code.
This PR hopes to help kill this problem by introducing a new /living subtype, /living/basic. The idea of this refactor is to slowly start moving all old simple_animals to this new system, moving over behaviors like charging and more extravagant mobs like megafauna over bit by bit similar to how newfood was implemented.
One of the other big goals of this refactor is to move many of the fringe simple animal behaviors into either AI datums, or components/elements. (Some of which still needs to be done in this PR).
As a proof of concept, I created the base mob/living/basic, and moved cockroaches over to the system. Since cockroaches have both a passive, melee and ranged mob.
This PR does slightly affect balance as the behavior isn't 1-on-1 due to it no longer running on the janky /hostile behavior, but I tried to keep the effects to a minimum, and the glockroach and hauberoach are not spawnable through many means as far as I know.
* The Failsafe can now recover from an deleted MC
Its also more reliable and can handle a situation where its main Loop runtimes and the MC is stuck
* Reset defcon level correctly
Oops left that in from debugging the levels
* Correctly recover SSasset
* Only decrease defcon if MC creation failed
Also add some sort sleep between emergency loops
* Makes the last two emergency actions manual procs
Since they are kinda unstantable its probalby best
if only admins call these manually
Its also more reliable and can handle a situation where its main Loop runtimes and the MC is stuck
You can also now debug Master/New()
While there will most likely never be any situation where the MC is just gone its still good to know that the game can recover from such a situation
For example maybe someone messed up a SDQL query or maybe someone wanted to delete the MC to create a new one hoping the Failsafe would do so for him
Repaths everything referring to "toxins" while actually meaning either the room in science or plasma gas. While this PR might be disrespectful to our forefathers, given this is (I believe) a holdover from as far back as the Exadv1 days, this has constantly irked me since I started working with the code. None of the player-facing stuff has referred to plasma as toxin since before 4407 hit, besides the Toxins Lab, and yet all of the type-paths are still pointing at toxins, making it a nightmare to search for in a map editor, and making the code needlessly easy to confuse with that of toxin damage. So this just fires it into the sun.
Anything relating to Toxins, the science subdepartment, now makes reference to Ordnance instead. This felt fitting enough given the focus of the subdepartment is around the creation of and testing of explosives.
Anything relating to plasma gas has, fittingly, been made to refer to plasma gas.
Edit: Ah yes, I feel I should probably apologise off the bat for the size of this PR- the code touched is mostly atmos machinery and simplemobs, a few sprites here and there, and of course the station maps + a few offstation maps.
Makes the code more legible and makes mapping less painful.
(The payment has been made)
Shouldn't have any visible effects. Atmos machines would occasionally complain about parent pipenets not existing during an update and this should resolve a category of those kinds of warning.
(This is a pain side effect of rebuilding being non instant, shoulda been dealt with a long time ago. Thank you -Lemon)
* Makes camera shaking more smooth
Makes shake_camera use more then instant steps for its animates. The concept is to not respect duration quite so
strictly. So instead of instantly moving, we get the distance to travel, pick the lower of the two numbers to
preserve some jerk, and multiply it by tile size * 0.7. So effectively 0.7 ticks per tile. I've found this works
reasonably well for making shaking events look less shit, though there is room to make it a preference option in
future, since this number could go between 0 for the old behavior, and say 5 for slow but not say, visually
hasurdus speeds.
* Makes large close range explosions shake more then far out ones
The old system for calculating explosion shake was sqrt(delta/10) when the delta could really only ever go up to
like 20. So it maxed out at 1.3 or so, which doesn't really fit being hit directly by a bomb. I hate old coders.
I'm changing the fomula to sqrt(range/(dist + 1)), which works out to roughly 4.5 at 20 near_distance, which is
a maxcap. Scales reasonably fast, 4 tiles or so, so this primarially effects getting hit full on with a big
bomb
Fixes a ton of harddels, sourced from #59996
I think this brings us down to like, ~100 per round from ~200, with only like 20 of those being proper hell failures. I've seen harddel profiles below 1 second of total cost. Feeling good.
See you on the other side
Makes the cryopod control computer into a weakref, never trust bee code
Converts brig door timer internal lists to weakrefs
Fixes a harddel caused by qdeling a motion sensitive camera after it had left its source area, jesus christ why didn't we do this already holy shit
Converts the radio implant ref held by the antenna mutation to weakrefs because it isn't reliably cleaned up, makes the radio implant actually qdel its fucking radio
Removes the target var from the throwing datum, it does literally nothing and just exists to cause harddels, mostly for the singularity
Fixes a cable harddel sourced from things that try to enter blueprints after smoothing, but before roundstart. IE, shuttles. Removes shuttles from the blueprints
Fixes emmisive blockers being added post qdel
Removes some manual ghosting from cryopods, I initially did this for harddel reasons, but I figured out a better fix for that. I'm now doing it because it's got this really strange logic for like "re-entering the game" that doesn't actually link to what the ghostize proc does. We should remove this at some point
Fixes robot hud objects harddeling due to hanging refs
Fixes buildmode related hanging refs, I'm coming for you admin team
Fixes a few instances of trying to add the forensics component post qdel, hhhhhhhhhhh
Fixes some split personality harddels/weirdness
Replaces a use of disconnect_duct with an init qdel hint, I suspect there's more issues with duct harddels, I've seen some odd logs about ahhh the area_contents list, but we can worry about that later
Makes teleporter targets into weakrefs, properly types them as /atom
Makes frequency devices into weakrefs
Makes cameras remove themselves from camera nets on Destroy
Makes tgui ui datums implement destroy, this means if I ever see one hang a ref to user or whatever, I know there's an error with calling close() properly. I've seen this harddel once, but not after this change so I assume there was some error with close(). IDK maybe this is a papering over? Would have to ask @stylemistake
I've seen logs of beartraps being in world post del, putting a return there just in case. The same is true of nerf darts, but I haven't really looked into that yet
Makes a shoe's ref to untying alerts a weakref, yes this is needed.
Moves clearing client_in_contents to the Login of the new mob. This prevents doing things like ghosting someone before a mob qdel causing harddels
Fixes a harddel set sourced from adding a status effect to a qdeleted thing. Is this an error? I'm honestly not sure.
Converts bsa code to weakrefs
Converts the partner var of heat exchangers to weakrefs
Converts camera assemblies to weakrefs
Fixes some dumb behavior with ammo casings and assuming you'll be on a turf post Destroy parent call
Fixes? merger related harddels, you were never cleared from your own members list, so origin objects would end up making a new list, creating harddels. Potential input from @ninjanomnom about the logic
Chasms store a static list of "falling atoms", which only exists for chasms that go somewhere else. This list wasn't being cleared of qdeleted objects, which is what happens when you fall in most chasms. Fixes this, and converts the list to weakrefs.
Fixes some runtimes in both sheet code, and the weather listener element. This is here because runtime spam made testing more of a pain, didn't think it needed its own pr
Fixes colorful reagent harddels sourced from reagents that were qdel'd before roundstart. I'm only like 50% sure this actually got it, but the issue may have been solved by #60174, so eh
Turns the nuke op antag datum's ref to the war button into a weakref
Fixes some holopad code that was not nulling refs all the time
Converts camera bugs to weakrefs, this was the result of the bug being "reworked" like 6 years back without taking the existing ref clearing into account. Whole item needs a redo, but this'll do for now.
Ensures that the both pulling and pullee refs are cleared on Destroy
The crew monitor held all users in a non clearing list, makes that list a weakref because I hate everything
Oh and I removed all sources of gas_mixture qdeletion, I'm kinda unsure on this since it's not technically supported, but any harddels from it might? indicate something going wrong with like, gas passing logic. I'd like @MrStonedOne's thoughts, since I trust him to call me an idiot if I'm wrong.
<!-- Please add a short description of why you think these changes would benefit the game. If you can't justify it in words, it might not be worth adding. -->
## Why it's not good for the game
I crashed sybil like 10 times to get this data, I'm gonna put it to good use. Don't think you're safe sybilites, I'm coming for you.
Maintenance loot spawners now fire at roundstart, rather than during the initialization of the Atoms subsystem.
Maintenance loot spawners do not qdel themselves during the spawning process, but persist, and merely hide themselves from living players.
As a cute side effect, this lets preround observers look at the colourful maintenance loot dice icons before the round starts. Observers can continue to see the loot dice post-round, but they are deliberately
very faint.
This queue allows things that hold refs to objects intermitently to free them on their own, without causing
harddels. A big part of this is byond, who will occasionally hold refs to an object for up to I think? 5
minutes. The old timer was lower for fear of memory leaks, but that's not as big a concern anymore
* New arrivals are less likely to spawn on top of each other.
This will close#58311 (Just an issue about new arrivals spawning on top of each other like fools)
Co-authored-by: coiax <yellowbounder@gmail.com>
Requested by oranges and inspired by the upcoming event. A new subsyetem, non-processing (for now), aimed at providing some toggle switches that can be flipped as a last ditch effort to save some CPU cycles by sacrificing some non-critical mechanics. Below you can see each individual toggle.
Screenshot of the admin panel:
image
Surely there are more opportunities for toggles I missed, but adding new ones is not very difficult at all.
Why It's Good For The Game
Better performance during extreme pop, I hope.
Changelog
cl
code: Introduces the Lag Switch subsystem for when a smoother experience is worth trading a few bells and whistles for. Performance enhancement measures can be togged by admins with the Show Lag Switches admin verb or enabled automatically at a pop amount set via config.
config: Added a new config var: number/auto_lag_switch_pop
Add a new simple station trait, that replaces everyone's ID with a
wallet containing said ID, as well as their starting money, and if
they're lucky, a little bonus surprise.
In addition, adds a force variable to station traits, which will
always enable them, rather than having to edit weight and probability
of rolling a trait of that sign (postive, negative).
Co-authored-by: Rohesie <rohesie@gmail.com>
Part of my document on duplicating verb panel functionality so at some point in the far, far future we may be rid of it. you can read more here.
hackmd.io/tJ5H3opRQiG1GCN1sP142w
The other thing the status panel needs to do is send time dilation to the tgui chat. Having messed around with tgui chat, I am very confident i'm going to need stylemistake or mothblocks help me out with that one, it's so beyond my level it isn't even funny.
I know this is a VERY long description for a very small change but expect more of this in the future
* Converts looping sounds from a list of play locations to just the one
* Updates all uses of looping sounds to match the new arg
* Adds an area based sound manager that hooks into looping sounds to drive the actual audio. I'll be using this to redo how weather effects handle sound
* Some structrual stuff to make everything else smoother
Timers now properly return the time left for client based timers
Weather sends global signals when it starts/stops
Looping sounds now use their timerid var for all their sound related timers, not just the main loop
* This is the painful part
Adds an area sound manager component, it handles the logic of moving into new areas potentially creating new
sound loops. We do some extra work to prevent stacking sound loops.
Adds an ash storm listener element that adds a tailored area sound manager to clients on the lavaland z level.
It's removed on logout.
Adds the ash_storm_sounds assoc list, a reference to this is passed into area sound managers, and it's modified
in a manner that doesn't break the reference in ash_storm (This is what I hate)
* Hooks ash storm listener into cliented mobs and possessed objects
* Documents the odd ref stuff, adds an ignore start var to looping sounds, fixes some errors and lint issues
* Applies kyler's review
banging
Co-authored-by: Kylerace <kylerlumpkin1@gmail.com>
* Cleans up some var names, reduces the amount of looping we do in some areas
* Makes the code compile, redoes the movement listener to be more general
* fuck
* We don't need to detach on del if we're just removing signals on detach
* Should? work
* if(direct) memes
Co-authored-by: Kylerace <kylerlumpkin1@gmail.com>
Redoes how logs are generated, calling profile() does not guarentee that the output logs will contain one of each entry, if the value is 0 it just doesn't return data for that entry. So we need to do some memes to make sure things are working properly
Removes a debug proc that I forgot about
cl
server: Added configs to disable laggy hard deletes once they lag the server too much.
admin: laggy hard deletes only output once per type path.
/cl
closes#58379
1 second was too long in the current lag market.
(postpone doesn't work here because of a bug i'll fix in another webedit in another pr since its actually a mc bug)
Unobvious problem spot
#define BUCKET_POS(timer) (((round((timer.timeToRun - SStimer.head_offset) / world.tick_lag)+1) % BUCKET_LEN)||BUCKET_LEN)
With tick_lag equal to 0.1, 0.25, 0.5, rounding of division is normal. But at other values it may be shifted either more or less due to the specifics of floating-point storage and processing. Numbers 0.1, 0.25, 0.5 have blank mantissa, unlike others which lead to numbers such as 245.0000004 when divided.
PS: tick_lag is rounded to the first two decimal places.
Fixes
Rewrote the circular doubly linked list to a regular doubly linked list, because it could cause timers to loop infinitely.
Fixed re-creation of a bucket if the timer does not have a callback.
Added debug stat indicator RST to MC SStimer.
Added optional ability to log dump all timers on crash.
Fixed subsystem logic when a bucket position is misplaced due to division and rounding inaccuracy. The system now captures such rounding errors and restores the correct timer position.
References
[RU] SS220 Paradise port process from TGstation and fixes:
ss220-space/Paradise#5ss220-space/Paradise#10ss220-space/Paradise#26ss220-space/Paradise#32ss220-space/Paradise#37
Contributors
@semoro: fixes
@Bizzonium: port
Changelog
cl Semoro and azizonkg
fix: Ported fixes of SStimer subsystem from RU SS220 Paradise
config: Added a new config var: flag/log_timers_on_bucket_reset
/cl
Refactors dbcore to work off a subsystem if executed async and limits the maximum amount of concurrent async queries to 25.
This has been tested locally on a mysql docker image and there were no crashes (as long as you didn't run it with debug extools) + data was getting recorded fine.
Why It's Good For The Game
May or may not resolve terry crashes, however, each query creates a new thread which takes up 2mb, preventing the game from using that 2mb. This can lead to ooms if they stack up, e.g. due to poor connectivity. This solves that issue.
maintainer note: this did not actually resolve the crashes, but has value anyway. Crashes were sidestepped fixed by finding out Large Address Awareness works
cl
refactor: Refactors dbcore.dm to possibly resolve the crashes that happen on Terry.
/cl
This makes it so during unit tests, adding a text based overlay to something will runtime if the icon does not have an icon state matching that text. I would do this during normal compiles as well but getting the icon states from an icon is surprisingly expensive.
Co-authored-by: LemonInTheDark <58055496+LemonInTheDark@users.noreply.github.com>
Converts most spans into span procs. Mostly used regex for this and sorted out any compile time errors afterwards so there could be some bugs.
Was initially going to do defines, but ninja said to make it into a proc, and if there's any overhead, they can easily be changed to defines.
Makes it easier to control the formatting and prevents typos when creating spans as it'll runtime if you misspell instead of silently failing.
Reduces the code you need to write when writing spans, as you don't need to close the span as that's automatically handled by the proc.
(Note from Lemon: This should be converted to defines once we update the minimum version to 514. Didn't do it now because byond pain and such)
