mirror of
https://github.com/Bubberstation/Bubberstation.git
synced 2026-07-11 16:14:08 +01:00
master
8 Commits
| Author | SHA1 | Message | Date | |
|---|---|---|---|---|
 MrMelbert
|
e0bdfc3f5f |
Dynamic Rework (#91290)
Implements https://hackmd.io/@tgstation/SkeUS7lSp , rewriting Dynamic
from the ground-up
- Dynamic configuration is now vastly streamlined, making it far far far
easier to understand and edit
- Threat is gone entirely; round chaos is now determined by dynamic
tiers
- There's 5 dynamic tiers, 0 to 4.
- 0 is a pure greenshift.
- Tiers are just picked via weight - "16% chance of getting a high chaos
round".
- Tiers have min pop ranges. "Tier 4 (high chaos) requires 25 pop to be
selected".
- Tier determines how much of every ruleset is picked. "Tier 4 (High
Chaos) will pick 3-4 roundstart[1], 1-2 light, 1-2 heavy, and 2-3
latejoins".
- The number of rulesets picked depends on how many people are in the
server - this is also configurable[2]. As an example, a tier that
demands "1-3" rulesets will not spawn 3 rulesets if population <= 40 and
will not spawn 2 rulesets if population <= 25.
- Tiers also determine time before light, heavy, and latejoin rulesets
are picked, as well as the cooldown range between spawns. More chaotic
tiers may send midrounds sooner or wait less time between sending them.
- On the ruleset side of things, "requirements", "scaling", and
"enemies" is gone.
- You can configure a ruleset's min pop and weight flat, or per tier.
- For example a ruleset like Obsession is weighted higher for tiers 1-2
and lower for tiers 3-4.
- Rather than scaling up, roundstart rulesets can just be selected
multiple times.
- Rulesets also have `min_antag_cap` and `max_antag_cap`.
`min_antag_cap` determines how many candidates are needed for it to run,
and `max_antag_cap` determines how many candidates are selected.
- Rulesets attempt to run every 2.5 minutes. [3]
- Light rulesets will ALWAYS be picked before heavy rulesets. [4]
- Light injection chance is no longer 100%, heavy injection chance
formula has been simplified.
- Chance simply scales based on number of dead players / total number
off players, with a flag 50% chance if no antags exist. [5]
[1] This does not guarantee you will actually GET 3-4 roundstart
rulesets. If a roundstart ruleset is picked, and it ends up being unable
to execute (such as "not enough candidates", that slot is effectively a
wash.) This might be revisited.
[2] Currently, this is a hard limit - below X pop, you WILL get a
quarter or a half of the rulesets. This might be revisited to just be
weighted - you are just MORE LIKELY to get a quarter or a half.
[3] Little worried about accidentally frontloading everything so we'll
see about this
[4] This may be revisited but in most contexts it seems sensible.
[5] This may also be revisited, I'm not 100% sure what the best / most
simple way to tackle midround chances is.
Other implementation details
- The process of making rulesets has been streamlined as well. Many
rulesets only amount to a definition and `assign_role`.
- Dynamic.json -> Dynamic.toml
- Dynamic event hijacked was ripped out entirely.
- Most midround antag random events are now dynamic rulesets. Fugitives,
Morphs, Slaughter Demons, etc.
- The 1 weight slaughter demon event is gone. RIP in peace.
- There is now a hidden midround event that simply adds +1 latejoin, +1
light, or +1 heavy ruleset.
- `mind.special_role` is dead. Minds have a lazylist of special roles
now but it's essentially only used for traitor panel.
- Revs refactored almost entirely. Revs can now exist without a dynamic
ruleset.
- Cult refactored a tiny bit.
- Antag datums cleaned up.
- Pre round setup is less centralized on Dynamic.
- Admins have a whole panel for interfacing with dynamic. It's pretty
slapdash I'm sure someone could make a nicer looking one.
- Maybe some other things.
See readme for more info.
Will you see a massive change in how rounds play out? My hunch says
rounds will spawn less rulesets on average, but it's ultimately to how
it's configured
🆑 Melbert
refactor: Dynamic rewritten entirely, report any strange rounds
config: Dynamic config reworked, it's now a TOML file
refactor: Refactored antag roles somewhat, report any oddities
refactor: Refactored Revolution entirely, report any oddities
del: Deleted most midround events that spawn antags - they use dynamic
rulesets now
add: Dynamic rulesets can now be false alarms
add: Adds a random event that gives dynamic the ability to run another
ruleset later
admin: Adds a panel for messing around with dynamic
admin: Adds a panel for chance for every dynamic ruleset to be selected
admin: You can spawn revs without using dynamic now
fix: Nuke team leaders get their fun title back
/🆑
(cherry picked from commit
|
||
|
Jacquerel
|
434e4e435e |
Removes Secondary & Final Objectives from Traitors (#89466)
## About The Pull Request  Pre-discussed with @Watermelon914, this PR removes Secondary & Final Objectives from all Traitors, rather than just midround ones. It also removes all of the surrounding supporting code. Randomly assigned Primary Objectives still exist, I just used the ability to rewrite mine to take the screenshot. In terms of final objectives, the surrounding items that were available still exist but don't necessarily have sources. If anyone has good ideas for readding these in some other form it can be done in future PRs. It also allows all traitors to buy the Contractor kit, previously limited to midround traitors which lacked secondary objectives, because now all traitors lack secondary objectives. This essentially limits all traitors to a maximum of 20 TC (16 if they spawn with an uplink implant). Currently I don't foresee that they strictly need any additional way of gaining TC during a round as 20 is quite sufficient, but it may take some time to adjust and get used to it after such a long time of having access to more. If we need to adjust the starting value or add a slow drip of more points over time or something, that can be done in followup PRs. This also removes the ability to recreate your uplink added by my beautiful wife in #74315 This was part of the progression traitor design document, but ultimately probably a bad idea as it essentially made traitors impossible to properly disarm. You will once more just need to carefully protect your uplink. **This does not remove the threat/progression system**. Like midround traitors, all Reputation requirements on gear are now simple timelocks, most of which will have elapsed by the time 30 minutes have passed. **Finally** this PR also adds Romerol to the traitor uplink for 25 TC and 30 minutes of reputation, as a treat (and because I removed the final objective that previously granted it). ## Why It's Good For The Game We've tried this system for a long time (3 years last month!) and while I think it had a lot of promise, enabled some cool moments, and also solved several of the problems it set out to solve, overall I think some of the behaviours it has encouraged in players have been overall negative for the game. While the _game systems_ are fine, even quite fun and cool (especially final objectives) I am of the opinion that having them in the game creates a net negative purely in the way that they react with players' _brains_, creating incentives towards behaviour we don't actually want people to pursue. While it's hard-to-impossible to prove any of this with hard data, there has been a prevailing feeling for some time among many (though certainly not all) people that the simple fact of _having_ a constant drip-feed of objective available to players leads directly to less interesting antagonist play. While certainly nobody is _forced_ to do secondary objectives you are directly and quite strongly rewarded for doing so, doing so efficiently, and doing so in a way which makes sure that nobody (alive) sees you do it. This leads to a tendency to play defensively and try to maximise the number of tasks you can complete in one round, which also has a knock-on effect of generally minimising the number of people you attempt to interact with in a round (unless you are killing them). Even people who _intend_ on doing some more interesting gimmick can fall into this trap, as "having more tools" is always useful for anyone who is intending on any kind of plan at all, but then executing on the secondary objectives again incentivises you to lay low, not interact with anyone, be efficient, and then reduces the time you are spending doing the thing that's your actual plan for the round. Removing the ever-present temptation to fish for extra TC leaves "doing whatever your actual plan is" as the sole thing to optimise. Final Objectives too have created unfortunate psychological effects between crewsided players and other antagonists. Because of the _threat_ (no matter how remote, Final Objectives have always been tuned to be appropriately rare) that leaving any antagonist alone will cause them to snowball by acquiring more power, it starts to feel foolish to respond to any threat with less than the maximum possible level of force even if they seem relatively innocuous in the moment. This even has an effect on other non-progression antagonists, as traitors are the most common antagonist type and how people treat them is going to be their default level of reaction to most other station threats. While there has always been the promise of expanding the system with novel and exciting objectives that leverage appearing mid-round to do something unique, we've taken very little advantage of that over time. Most objectives we have added that didn't boil down to "kill someone, with a twist" have been somewhat unsuccessful, serving either as ways to get yourself arrested and killed for no reason or ways to get free telecrystals by doing something the crew don't really care about stopping you from doing. The option still exists to add more roundstart objectives to traitors, if someone suddenly has a great idea that would fit in this space. The ideal outcome of making this change is a slight relaxation of crew attitude towards feeling like their only option after catching an antagonist that isn't sandbagging is to permanently remove them from the round (although it's fine to do this still in many scenarios), and a broadening of traitorous activity which is not purely focused on collecting as many checkboxes as possible and might give people more time to roleplay with other players, not worrying that this time could have been more efficiently spent pursuing a different secondary goal. I don't anticipate or desire that this will prevent traitors from killing anyone (or even stop them from killing people they don't have a specific objective to kill), I just want to remove the FOMO from people's minds. Also this gives us something to talk about at the coder townhall meeting on the 22nd. ## Changelog 🆑 del: Misplaced or stolen traitor uplinks can no longer be recreated using a radio code and special device, guard yours carefully or buy a backup implant. del: Roundstart traitors can no longer take on additional objectives in order to earn additional Telecrystals and fast-forward any unlock timers on items. They also cannot earn the ability to complete a Final Objective. balance: Roundstart traitors can now buy the Contractor Kit from their traitor uplink, rather than only midround traitors. add: Traitors can buy Romerol for 25 TC, after 30 minutes of time has passed in a round. /🆑 |
||
|
Timberpoes
|
6808a082eb |
Assorted changes to job assignment code and logging. Runtime free, guaranteed or your money back. Price: $£0. (#85947)
## About The Previous Pull Request #85308 reverted by #85929  ~~Causes the round to not start when a player isn't eligible for any jobs at a specific priority level due to runtimes trying to `pick()` from an empty list aborting the entire job assignment stack.~~ (Fixed???? by https://github.com/tgstation/tgstation/pull/85947/commits/e0e9f2f430079d4ab7097abe12e75f934131a638) Maybe we should test merge this for a mo just to make sure no more cheeky runtimes pop up before merging. ## About The Pull Request This PR does a couple of minor things: Makes the job debug logging a bit easier to follow. Minorly brings some SSjob code up to code standards, converting proc names to snake_case and doing some otherm is cleanup. Refactored some stuff into different procs, updated some comments. And some major things: Changes the job assignment logic. Old behaviour > Assign dynamic priority roles > Force one Head of Staff (if possible) > Assign all AIs > Assign overflow roles (bugged in 2 ways) > Shuffle the available jobs list once, at the start of the random job assignment loop > Pick and assign random jobs for random players from High prefs down, with a priority on Head of Staff roles > Handle everyone that couldn't be assigned a random job New behaviour > Assign dynamic priority roles > Assign all Head of Staff roles to players with High prefs > If no Head of Staff was made in the above way, force one Head of Staff (if possible) > Assign all AIs > Assign overflow roles (fixed) > Prioritise and fill unfilled head roles at each job priority pref level, from High prefs down. > Build a list of all jobs that each unassigned player could be eligible for at the above pref level. > Pick a job from that list at random and assign it to the player. > Handle everyone that couldn't be assigned a random job. In reality there should be little impact on overall job assignment, the code changes read more as semantics. For example, the priority check for filling Head slots will have the same candidate pool in both old and new versions, but in the new version we're more clearly saying that Heads are important and we want to prioritise filling them for the sake of round progression even though the outcome in new and old is the same. A key change will lead to an increase in assistants - Overflow fixes. Currently the code block to do early assignments to the Overflow role doesn't work - or works but not as you'd expect. The idea was is that because enabling the Overflow role in the prefs menu is an On/Off toggle that sets the job to High priority when enabled and prevents any other High priority pref, players that have the Overflow role enabled will **always** get it. It's their highest priority job with infinite slots. So we do a pass right at the start to give everyone with the Overflow role enabled that role and save us wasting time later on in random job code giving them that same role but with more work. The problem is the code for this only assigns the Overflow role to people with it set to Low priority in their prefs, resulting in log readouts like: ``` [2024-07-27 09:49:43.469] DEBUG-JOB: DO, Running Overflow Check 1 [2024-07-27 09:49:43.469] DEBUG-JOB: Running FOC, Job: /datum/job/assistant, Level: Low Priority [2024-07-27 09:49:43.472] DEBUG-JOB: FOC player job enabled at wrong level, Player: Radioprague, TheirLevel: Medium Priority, ReqLevel: Low Priority [2024-07-27 09:49:43.472] DEBUG-JOB: FOC player job enabled at wrong level, Player: Caluan, TheirLevel: High Priority, ReqLevel: Low Priority [2024-07-27 09:49:43.473] DEBUG-JOB: FOC player job enabled at wrong level, Player: Caractaser, TheirLevel: High Priority, ReqLevel: Low Priority [2024-07-27 09:49:43.473] DEBUG-JOB: FOC player job enabled at wrong level, Player: Apsua, TheirLevel: High Priority, ReqLevel: Low Priority [2024-07-27 09:49:43.475] DEBUG-JOB: FOC player job enabled at wrong level, Player: Bebrus2, TheirLevel: Medium Priority, ReqLevel: Low Priority [2024-07-27 09:49:43.475] DEBUG-JOB: AC1, Candidates: 0 ``` Where nobody gets pre-assigned the overflow role because their prefs are all set to the High priority from being toggled... Except wait a second, some people have it at Medium priority when it should just be a No Role/High Priority Role toggle? And herein we meet a problem. My hypothesis is that traits and stuff that change the overflow have allowed players to set the "ordinary" overflow role of Assistant to Medium and/or Low priority. This still shows as enabled in the prefs menu, but leads to an outcome where a player with assistant enabled is assigned Cook instead. ``` [2024-07-27 09:49:47.775] DEBUG-JOB: DO, Running Overflow Check 1 [2024-07-27 09:49:47.775] DEBUG-JOB: Running FOC, Job: /datum/job/assistant, Level: Low Priority ... [2024-07-27 09:49:43.475] DEBUG-JOB: FOC player job enabled at wrong level, Player: Bebrus2, TheirLevel: Medium Priority, ReqLevel: Low Priority ... [2024-07-27 09:49:47.987] DEBUG-JOB: Running AR, Player: Bebrus2, Job: /datum/job/cook, LateJoin: 0 ``` So players with the Overflow job pref set to Low (an unexpected state, should be disabled or High) would be guaranteed to get that role if none of the higher priority Head of Staff/AI/Dynamic roles took over via the bugged "force overflow for people with the pref enabled" proc. Players with the Overflow job pref set to High would be guaranteed to get that role if none of the higher priority Head of Staff/AI/Dynamic roles took over via the random job assignment code giving them their Highest priority role thanks to the infinite job slots of the Overflow. And players with the Overflow job pref set to Medium (an unexpected state, should be disabled or High) would get Assistant if the shuffle step of the available jobs list put Assisstant before any of the other jobs they had prefs enabled for at Medium that weren't already filled, otherwise they'd get another random job. This code is now changed to ignore the priority the player has set when looking for people to fill the overflow role. As long as it **is** enabled, the player will get it unless they're forced into a dynamic ruleset role (AI when malf rolls) or a Head of Staff role due to their other prefs (they have RD set to med or low, and no other player has a Head of Staff at high so they get randomly picked and miss the overflow role). This will increase the number of assistants in shifts where their pref state has Assisstant in the bugged Medium priority, but doesn't change it for bugged Low and not-bugged High/On priority. On the other side of the coin, we have how the random jobs are picked. They're kinda not random, and I noticed this reading the logs then reading the code. The list of available jobs to pick from is randomly shuffled - but only **once**. All players pull from a list of jobs in the same order. So you end up with a log block like this: ``` [2024-07-27 09:49:47.985] DEBUG-JOB: DO pass, Player: Pierow, Level:3, Job:Botanist [2024-07-27 09:49:47.985] DEBUG-JOB: Running AR, Player: Pierow, Job: /datum/job/botanist, LateJoin: 0 [2024-07-27 09:49:47.985] DEBUG-JOB: Player: Pierow is now Rank: Botanist, JCP:0, JPL:2 [2024-07-27 09:49:47.986] DEBUG-JOB: DO pass, Player: Daddos, Level:3, Job:Botanist [2024-07-27 09:49:47.986] DEBUG-JOB: Running AR, Player: Daddos, Job: /datum/job/botanist, LateJoin: 0 [2024-07-27 09:49:47.986] DEBUG-JOB: Player: Daddos is now Rank: Botanist, JCP:1, JPL:2 [2024-07-27 09:49:47.986] DEBUG-JOB: FOC job filled and not overflow, Player: Bebrus2, Job: /datum/job/botanist, Current: 2, Limit: 2 [2024-07-27 09:49:47.987] DEBUG-JOB: FOC player job not enabled, Player: Bebrus2 [2024-07-27 09:49:47.987] DEBUG-JOB: DO pass, Player: Bebrus2, Level:3, Job:Cook [2024-07-27 09:49:47.987] DEBUG-JOB: Running AR, Player: Bebrus2, Job: /datum/job/cook, LateJoin: 0 [2024-07-27 09:49:47.988] DEBUG-JOB: Player: Bebrus2 is now Rank: Cook, JCP:0, JPL:1 [2024-07-27 09:49:47.988] DEBUG-JOB: FOC player job not enabled, Player: Redwizz [2024-07-27 09:49:47.988] DEBUG-JOB: FOC job filled and not overflow, Player: Redwizz, Job: /datum/job/cook, Current: 1, Limit: 1 ``` The list is shuffled into an order of something like `list("Scientist", "Botanist", "Cook", "Sec Officer", ...)` then iterated over for each player. So every random job selection goes: > "Does Player1 have Scientist enabled and at the right priority? No? Okay, Botanist? Yes? You get botanist." > "Does Player2 have Scientist enabled and at the right priority? No? Okay, Botanist? Yes? You get botanist." > "Does Player3 have Scientist enabled and at the right priority? No? Okay, Botanist has no slots left so we'll remove it from the list. Okay, Cook? Yes? You get cook." > "Does Player4 have Scientist enabled and at the right priority? No? Okay, Cook has no slots left so we'll remove it from the list. Okay, Sec Officer? ..." This can lead to stacked individual departments if it gets randomly rolled to the start of the list in the shuffle, and completely empty departments if they end up at the end. On high pop shifts this is probably less of an issue. Player prefs add noise to this and as departments at the front fill up, those at the back pick up some of the lower pref players. But have you ever had a shift where there's just like... No fucking sec even though there's tons of players? The logging (before I made changes in this PR) was a bit ass, but my hypothesis there is that sec officer was shuffled right at the end of the random job list, so every other department was filled up before sec officers were picked. To mitigate this, I made the list shuffle every single time the game picks a random available job for the player. This should lead to a more balanced selection of available jobs by avoiding situations where the code is biased towards packing some departments by accident. ## Why It's Good For The Game Overflow fixes mean people who go to their prefs and see the Overflow Role is On will all have the same experience - They will be the Overflow role. More random random job selection should prevent individual departments having a jobs be stacked when it would have otherwise been possible for a more balanced selection but the code unintentially biased random departments to be overstaffed and understaffed each shift. ## Changelog 🆑 fix: Having the Overflow Role set to On will properly ensure you get that role at a High priority as intended by the game code. fix: Job selection is now a little bit more random. Fixes an unintentional bias in random job assignment that could lead to feast-or-famine for roles where everyone is assigned one job and nobody is assigned another job. /🆑 |
||
|
Kyle Spier-Swenson
|
4d1639b04c |
Revert "Assorted changes to job assignment code and logging." (#85929)
Reverts tgstation/tgstation#85308  |
||
|
Timberpoes
|
1eef540054 |
Assorted changes to job assignment code and logging. (#85308)
## About The Pull Request
This PR does a couple of minor things:
Makes the job debug logging a bit easier to follow.
Minorly brings some SSjob code up to code standards, converting proc
names to snake_case and doing some otherm is cleanup.
Refactored some stuff into different procs, updated some comments.
And some major things:
Changes the job assignment logic.
Old behaviour
> Assign dynamic priority roles
> Force one Head of Staff (if possible)
> Assign all AIs
> Assign overflow roles (bugged in 2 ways)
> Shuffle the available jobs list once, at the start of the random job
assignment loop
> Pick and assign random jobs for random players from High prefs down,
with a priority on Head of Staff roles
> Handle everyone that couldn't be assigned a random job
New behaviour
> Assign dynamic priority roles
> Assign all Head of Staff roles to players with High prefs
> If no Head of Staff was made in the above way, force one Head of Staff
(if possible)
> Assign all AIs
> Assign overflow roles (fixed)
> Prioritise and fill unfilled head roles at each job priority pref
level, from High prefs down.
> Build a list of all jobs that each unassigned player could be eligible
for at the above pref level.
> Pick a job from that list at random and assign it to the player.
> Handle everyone that couldn't be assigned a random job.
In reality there should be little impact on overall job assignment, the
code changes read more as semantics. For example, the priority check for
filling Head slots will have the same candidate pool in both old and new
versions, but in the new version we're more clearly saying that Heads
are important and we want to prioritise filling them for the sake of
round progression even though the outcome in new and old is the same.
A key change will lead to an increase in assistants - Overflow fixes.
Currently the code block to do early assignments to the Overflow role
doesn't work - or works but not as you'd expect. The idea was is that
because enabling the Overflow role in the prefs menu is an On/Off toggle
that sets the job to High priority when enabled and prevents any other
High priority pref, players that have the Overflow role enabled will
**always** get it. It's their highest priority job with infinite slots.
So we do a pass right at the start to give everyone with the Overflow
role enabled that role and save us wasting time later on in random job
code giving them that same role but with more work.
The problem is the code for this only assigns the Overflow role to
people with it set to Low priority in their prefs, resulting in log
readouts like:
```
[2024-07-27 09:49:43.469] DEBUG-JOB: DO, Running Overflow Check 1
[2024-07-27 09:49:43.469] DEBUG-JOB: Running FOC, Job: /datum/job/assistant, Level: Low Priority
[2024-07-27 09:49:43.472] DEBUG-JOB: FOC player job enabled at wrong level, Player: Radioprague, TheirLevel: Medium Priority, ReqLevel: Low Priority
[2024-07-27 09:49:43.472] DEBUG-JOB: FOC player job enabled at wrong level, Player: Caluan, TheirLevel: High Priority, ReqLevel: Low Priority
[2024-07-27 09:49:43.473] DEBUG-JOB: FOC player job enabled at wrong level, Player: Caractaser, TheirLevel: High Priority, ReqLevel: Low Priority
[2024-07-27 09:49:43.473] DEBUG-JOB: FOC player job enabled at wrong level, Player: Apsua, TheirLevel: High Priority, ReqLevel: Low Priority
[2024-07-27 09:49:43.475] DEBUG-JOB: FOC player job enabled at wrong level, Player: Bebrus2, TheirLevel: Medium Priority, ReqLevel: Low Priority
[2024-07-27 09:49:43.475] DEBUG-JOB: AC1, Candidates: 0
```
Where nobody gets pre-assigned the overflow role because their prefs are
all set to the High priority from being toggled... Except wait a second,
some people have it at Medium priority when it should just be a No
Role/High Priority Role toggle?
And herein we meet a problem. My hypothesis is that traits and stuff
that change the overflow have allowed players to set the "ordinary"
overflow role of Assistant to Medium and/or Low priority.
This still shows as enabled in the prefs menu, but leads to an outcome
where a player with assistant enabled is assigned Cook instead.
```
[2024-07-27 09:49:47.775] DEBUG-JOB: DO, Running Overflow Check 1
[2024-07-27 09:49:47.775] DEBUG-JOB: Running FOC, Job: /datum/job/assistant, Level: Low Priority
...
[2024-07-27 09:49:43.475] DEBUG-JOB: FOC player job enabled at wrong level, Player: Bebrus2, TheirLevel: Medium Priority, ReqLevel: Low Priority
...
[2024-07-27 09:49:47.987] DEBUG-JOB: Running AR, Player: Bebrus2, Job: /datum/job/cook, LateJoin: 0
```
So players with the Overflow job pref set to Low (an unexpected state,
should be disabled or High) would be guaranteed to get that role if none
of the higher priority Head of Staff/AI/Dynamic roles took over via the
bugged "force overflow for people with the pref enabled" proc.
Players with the Overflow job pref set to High would be guaranteed to
get that role if none of the higher priority Head of Staff/AI/Dynamic
roles took over via the random job assignment code giving them their
Highest priority role thanks to the infinite job slots of the Overflow.
And players with the Overflow job pref set to Medium (an unexpected
state, should be disabled or High) would get Assistant if the shuffle
step of the available jobs list put Assisstant before any of the other
jobs they had prefs enabled for at Medium that weren't already filled,
otherwise they'd get another random job.
This code is now changed to ignore the priority the player has set when
looking for people to fill the overflow role. As long as it **is**
enabled, the player will get it unless they're forced into a dynamic
ruleset role (AI when malf rolls) or a Head of Staff role due to their
other prefs (they have RD set to med or low, and no other player has a
Head of Staff at high so they get randomly picked and miss the overflow
role).
This will increase the number of assistants in shifts where their pref
state has Assisstant in the bugged Medium priority, but doesn't change
it for bugged Low and not-bugged High/On priority.
On the other side of the coin, we have how the random jobs are picked.
They're kinda not random, and I noticed this reading the logs then
reading the code.
The list of available jobs to pick from is randomly shuffled - but only
**once**. All players pull from a list of jobs in the same order. So you
end up with a log block like this:
```
[2024-07-27 09:49:47.985] DEBUG-JOB: DO pass, Player: Pierow, Level:3, Job:Botanist
[2024-07-27 09:49:47.985] DEBUG-JOB: Running AR, Player: Pierow, Job: /datum/job/botanist, LateJoin: 0
[2024-07-27 09:49:47.985] DEBUG-JOB: Player: Pierow is now Rank: Botanist, JCP:0, JPL:2
[2024-07-27 09:49:47.986] DEBUG-JOB: DO pass, Player: Daddos, Level:3, Job:Botanist
[2024-07-27 09:49:47.986] DEBUG-JOB: Running AR, Player: Daddos, Job: /datum/job/botanist, LateJoin: 0
[2024-07-27 09:49:47.986] DEBUG-JOB: Player: Daddos is now Rank: Botanist, JCP:1, JPL:2
[2024-07-27 09:49:47.986] DEBUG-JOB: FOC job filled and not overflow, Player: Bebrus2, Job: /datum/job/botanist, Current: 2, Limit: 2
[2024-07-27 09:49:47.987] DEBUG-JOB: FOC player job not enabled, Player: Bebrus2
[2024-07-27 09:49:47.987] DEBUG-JOB: DO pass, Player: Bebrus2, Level:3, Job:Cook
[2024-07-27 09:49:47.987] DEBUG-JOB: Running AR, Player: Bebrus2, Job: /datum/job/cook, LateJoin: 0
[2024-07-27 09:49:47.988] DEBUG-JOB: Player: Bebrus2 is now Rank: Cook, JCP:0, JPL:1
[2024-07-27 09:49:47.988] DEBUG-JOB: FOC player job not enabled, Player: Redwizz
[2024-07-27 09:49:47.988] DEBUG-JOB: FOC job filled and not overflow, Player: Redwizz, Job: /datum/job/cook, Current: 1, Limit: 1
```
The list is shuffled into an order of something like `list("Scientist",
"Botanist", "Cook", "Sec Officer", ...)` then iterated over for each
player. So every random job selection goes:
> "Does Player1 have Scientist enabled and at the right priority? No?
Okay, Botanist? Yes? You get botanist."
> "Does Player2 have Scientist enabled and at the right priority? No?
Okay, Botanist? Yes? You get botanist."
> "Does Player3 have Scientist enabled and at the right priority? No?
Okay, Botanist has no slots left so we'll remove it from the list. Okay,
Cook? Yes? You get cook."
> "Does Player4 have Scientist enabled and at the right priority? No?
Okay, Cook has no slots left so we'll remove it from the list. Okay, Sec
Officer? ..."
This can lead to stacked individual departments if it gets randomly
rolled to the start of the list in the shuffle, and completely empty
departments if they end up at the end.
On high pop shifts this is probably less of an issue. Player prefs add
noise to this and as departments at the front fill up, those at the back
pick up some of the lower pref players.
But have you ever had a shift where there's just like... No fucking sec
even though there's tons of players? The logging (before I made changes
in this PR) was a bit ass, but my hypothesis there is that sec officer
was shuffled right at the end of the random job list, so every other
department was filled up before sec officers were picked.
To mitigate this, I made the list shuffle every single time the game
picks a random available job for the player. This should lead to a more
balanced selection of available jobs by avoiding situations where the
code is biased towards packing some departments by accident.
## Why It's Good For The Game
Overflow fixes mean people who go to their prefs and see the Overflow
Role is On will all have the same experience - They will be the Overflow
role.
More random random job selection should prevent individual departments
having a jobs be stacked when it would have otherwise been possible for
a more balanced selection but the code unintentially biased random
departments to be overstaffed and understaffed each shift.
## Changelog
🆑
fix: Having the Overflow Role set to On will properly ensure you get
that role at a High priority as intended by the game code.
fix: Job selection is now a little bit more random. Fixes an
unintentional bias in random job assignment that could lead to
feast-or-famine for roles where everyone is assigned one job and nobody
is assigned another job.
/🆑
---------
Co-authored-by: san7890 <the@san7890.com>
|
||
|
Tastyfish
|
bca463316b |
Makes integration test results be in color and have annotations (#66649)
About The Pull Request
Separated compiling the integration tests and running them as separate steps for organization purposes.
Added a TEST_ASSERT_NULL(value, reason) and TEST_ASSERT_NOTNULL(value, reason) because those are conceptually simple tests.
Makes the PASS and FAIL prefixes in the integration test log be green and red for better readability.
Failure reasons now display the file and line number.
In order to achieve this, direct calls to Fail() are now wrapped in a macro, TEST_FAIL(), as Fail() itself needs preprocessor stuff passed to it.
In the midst of updating it, I noticed multiple cases of tests directly calling Fail() and returning when they should have used a better macro, so those were updated. There was at least one case where it appeared that the code assumed that the test ended at Fail(), but made no attempt to do so, such as with the RCD test.
Feel free to double check all of the changed unit tests in case I made a functional behavior change, but they currently pass.
To take advantage of the previous change, failures are now marked as annotations. Note that outside of github, this creates an ugly-looking line but the primary environment is as a github action.
Examples with intentionally botched unit test:
image
image
image
Why It's Good For The Game
Makes inspecting failed unit tests significantly easier.
Changelog
N/A
|
||
|
Rohesie
|
77c9485ecb |
Cleans Job lists (hardcoded and not) (#64596)
* Removes hardcoded job from SSjkobs. "station_jobs", "head_of_staff_jobs", "additional_jobs_with_icons", "centcom_jobs" have been removed and replaced with flags and vars on the trims. |
||
|
Watermelon914
|
8fd85e9666 |
[MDB IGNORE] BIDDLE TRAITORS - Adds progression traitors. Refactors uplink code in its entirety (#63588)
Co-authored-by: Watermelon914 <3052169-Watermelon914@users.noreply.gitlab.com> Co-authored-by: tralezab <40974010+tralezab@users.noreply.github.com> Co-authored-by: Ghilker <42839747+Ghilker@users.noreply.github.com> Co-authored-by: tgstation-server <tgstation-server@tgstation13.org> Co-authored-by: gbfree <guillaumebfree@gmail.com> Co-authored-by: Sealed101 <75863639+Sealed101@users.noreply.github.com> Co-authored-by: Changelogs <action@github.com> Co-authored-by: John Willard <53777086+JohnFulpWillard@users.noreply.github.com> Co-authored-by: vincentiusvin <54709710+vincentiusvin@users.noreply.github.com> Co-authored-by: oranges <email@oranges.net.nz> Co-authored-by: Seth Scherer <supernovaa41@gmx.com> Co-authored-by: Jeremiah <42397676+jlsnow301@users.noreply.github.com> Co-authored-by: Azarak <azarak10@gmail.com> Co-authored-by: RandomGamer123 <31096837+RandomGamer123@users.noreply.github.com> Co-authored-by: LemonInTheDark <58055496+LemonInTheDark@users.noreply.github.com> Co-authored-by: Kylerace <kylerlumpkin1@gmail.com> Co-authored-by: ike709 <ike709@users.noreply.github.com> Co-authored-by: Iamgoofball <iamgoofball@gmail.com> Co-authored-by: MrMelbert <51863163+MrMelbert@users.noreply.github.com> Co-authored-by: Pepsilawn <reisenrui@gmail.com> Co-authored-by: AnturK <AnturK@users.noreply.github.com> Co-authored-by: Mothblocks <35135081+Mothblocks@users.noreply.github.com> Co-authored-by: SuperNovaa41 <supernovaa41@protonmail.com> Co-authored-by: GoldenAlpharex <58045821+GoldenAlpharex@users.noreply.github.com> Co-authored-by: Fikou <23585223+Fikou@users.noreply.github.com> Co-authored-by: san7890 <34697715+san7890@users.noreply.github.com> Co-authored-by: MMMiracles <lolaccount1@hotmail.com> Co-authored-by: Krysonism <49783092+Krysonism@users.noreply.github.com> Co-authored-by: Aziz Chynaliev <azizonkg@gmail.com> Co-authored-by: Tastyfish <crazychris32@gmail.com> Co-authored-by: Mooshimi <85910816+Mooshimi@users.noreply.github.com> Co-authored-by: carshalash <carshalash@gmail.com> Co-authored-by: Ryll Ryll <3589655+Ryll-Ryll@users.noreply.github.com> |