Just a hidden scratchpad.

Edit me
  • Investigate audio delay on Windows
    • On a second thought, not as bad as I expected
      • Was probably due to other applications taking control of the sound
      • Exiting other apps and installing nahimic maybe helped
    • Linux is still leagues ahead but w/e
  • Zoom semantics (for the last time)
    • FoW and FoV are different things
      • Fog of war is occlusion
      • Field of view determines the viewing size
        • Obviously fog of war is clipped by the field of view
        • But field of view settings are still effective even if fog of war is disabled
    • Maps can specify both FoV size and if FoW is enabled
    • There’s also FoW angle and FoW occlussion
      • Technically could switch both
  • arbitrary upscaling (only UI left)

    • verdict: use zoom on gameplay, let’s render GUI to a separate fbo and stretch it

    • as for the game literally nothing else is required except zoom=2

    • will partially solve the resolution advantage problem
    • note that we’ll have to adjust UI positions too, just zooming in won’t do
      • fastest will be to stretch it with an fbo, as we won’t have to rebuild ui fonts
      • do we really have to use an fbo though if we already virtually render it in 1920x1080?
      • what about mouse position?
    • will have to first scale the get_window_size value
  • Fist fighting could work by spawning an invisible fist melee item but only during attack
    • What if we have an invisible item without space etc that is only visible to melee system and drawer?
      • but you also need to infer collider or no?
  • Default spawn generation for FFA-like modes
    • later, GunGame will be team-based
  • The “Race” mode for surf maps could count how fast someone entered the final portal
    • E.g. 12000 entering velocity = +12 score, 1000 velocity = +1 score etc
    • This way players could compete within a single loop

  • reverb area

  • Do not hold name map in solvable
    • Pointless to sync over the network
    • serves no purpose logic-side as entity ids can serve identification
    • Only callouts should have it
      • however callouts can just have a flavour allocated
    • Alright maybe specific_names will come in handy
      • Right now it’s only used to map entity ids to nicknames
      • But we shouldn’t draw it like that
      • Yeah but let’s leave it like that for now, no time for this

  • black stroke for stamina/health circular bars because they’re invisible on brighter maps

  • Show a tooltip with some basic info when hovering resources in filesystem view like the initial size e.g. 32x64
    • length for sounds etc

  • Melee clash should: a) happen more often; b) and cancel damage

  • Merging static colliders
    • Purpose
      • Performance: Less overhead for visibility system
        • Although lights filter duplicate vertices, there would still be less raycasts if bodies were merged
      • Physics precision: a wall-touching character will always hit the edges between two bodies, pretty annoying
        • Weren’t much of a problem until now so it won’t be important until next MVP I guess
    • Instead of a full-blown geometry editor, let’s auto-detect static colliders with identical materials that touch each other
      • When should it be done?
        • We thought of post-inference in the cosmos but you have no guarantee that some constituent entities won’t be reinferred on their own for some reason (e.g. there could be some script)
        • Therefore it should be done on Editor->Cosmos conversion stage
          • There we have more civilised ways of determining whether two nodes can be truly merged
    • and then let the user compose complex shapes with just the primitives they come up with
      • That way we don’t force precise geometry editing on the mapper
      • and this is even easier on the mapper
  • Enter while filtering with ImGuiTextFilter could always select the target
    • E.g. go to node on scene
    • But leave this for later as it might be too automatic of a behavior
      • We never even used this filtering field before, people will just click on scene mostly
  • We might need to rework the lights to use two radiuses
    • But once we’re able to recreate cyberaqua etc
    • Yeah I think a total rework for the light system will be necessary
  • There’s no point making workarounds now because we’ll replace it with two 50%/100% falloff parameters
    • and this will solve pretty much everything
  • We’ll just tell people to place lights without changing defaults
    • Except for radius because it can stay like this
    • We might add half_radius too
      • More intuitive than 50%_falloff and anyways we cant put % in the prop name
  • Yeah for now let’s just disable editing of attenuations altogether
    • Maybe we’ll even end up with just this model
    • The rest works nicely, it’s only important that we can edit radius
      • and cutoff alpha for rudimentary “intensity” scaling
  • Make throwing headshots more consistent by considering e.g. the center of the rattle, the velocity of the rattle and head center
    • Actually not the center of the rattle
      • Whenever a bullet/thrown item is spawned, register current trajectory
        • Except for homing missiles, not sure how headshots should be handled here, the old way could work
          • Actually it could work - if it’s homing (so changed trajectory) never register the headshot, only on direct hit
    • All headshot/ricochet detection will benefit from it because there’s no measure that is more accurate
      • Except for ricochets on the homing missiles, we should calculate it by the current velocity etc
        • Even though realistically once it starts to “home” there won’t be any ricochets remaining
    • On ricochet, we’ll just have to update the trajectory
  • Entity types to add later once we won’t break abi
    • CALLOUT WITH ITS NAME IN THE COMPONENT!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
    • Plain sprited bodies but with arbitrary polygons per-entity, obviously
    • Dynamic decorations but non-organism like, we should have a specific entity for an organism
      • Pointless for them to be processed just like static animations
  • Filters window
    • Visible
      • (Checkboxes) Backgrounds Physical Foregrounds
    • Selectable
      • (Checkboxes) Backgrounds Physical Foregrounds

  • When resizing, might be a good idea to determine edges based on total aabb of selected entities

  • honor playtesting spawns

  • stronger highlight for icons? hover/activated is almost invisible

  • Grenades could/should have very low restitution against player bodies
  • Laser should switch to the knife/grenade being thrown when it is done with a gun in another hand
    • Actually just disable the second laser then because there are both now

Editor final MVP (post MVP) once we can ditch old maps

  • Deprecate ruleset maps.
    • Rebuild arena when changing mode
    • This will be necessary anyway with modes that can alter rebuild behavior (like enabling destroyed environments for zombie mode)
    • Then pick only one mode

  • remove fy_minilab_reloaded-only clause from project selector where it’s scanning for official areas

  • Ensure all official floors have some footsteps set

  • Full illumination maybe per entity too for foregrounds
    • Otherwise we’d need two types of ferns because cyberaqua has a full illumination one
  • wtf, warmup_theme shouldn’t be a flavour id but a sound effect input
    • It actually needs to be a flavor because it’s synchronized through the network so it needs to be an entity
      • it can’t be a sound effect
    • However later we’ll just allow sound decorations to be entities, pointless to have invariants here (?)
      • Remember it’s sent through the network in that case, although it’s just a few bytes
    • We’ll need to entitize at least gain and pitch to be able to honor these modifiers in sound node and theme properties
      • Well, fortunately we’ll easily be able to do this

  • search for TODO in editor generally

  • Markers should have letters per-entity just like factions
    • we can only set them per-invariant right now, there are bombsite_dummy_b etc, this is bad obv
  • Renames for weapons!!! We should really decide once and for all about some names so that they’re future-proof
    • Best to avoid two-word names except in rare cases
    • We can use some polish names because they’ll be sounding interesting to foreign players
    • elon_hrl -> rocket_launcher (we though
      • Good PR because it’s then obvious to anyone new to the game
      • Warx fq12 -> warx (there’s no other single-word intuitive name for an automatic shotgun)
      • datum_gun -> datum
      • ao44 -> revolver
        • actually let’s leave it at ao44 because we’ll use “Revolver” for something stronger
      • hpsr -> awka
      • lewsII -> lews (there’s no I anyway)
      • amplifier_arm -> amplifier

  • Sounds should have per-entity sound effect modifiers (we implement it in editor already)

  • Resizing animations

  • Watch out for MAX_IMAGES_IN_INTERCOSM
    • Let’s not allocate this statically
  • Check what happens when we overflow the texture size
    • with some giant gif for example
    • we can later implement the optimizations like separating atlases
  • AIR_DUCT -> VENT in official resources
    • We can’t do it now because it’s gonna break maps
  • Let’s make sprite’s special effect a boolset
    • and not just have one effect speed multiplier
    • for now one will just implicitly override another
  • Move render invariant to component, instead of allocating flavors for each override
    • WE STILL NEED AN OVERRIDE FOR physical->non physical.
    • Entire render component could fit into two bytes
    • Will even be faster to calculate render layer because won’t have to refer to flavor
    • When we componentize render, we can still have editable defaults in the resource

Others

  • Reset properties to default under Right-click
    • Could be easy but arguably not needed right now
    • Most of node properties won’t have the concept of default
    • Resources, maybe, still they have the reset button so one can always check and undo

  • Review the TODO: comments especially inside editor sources once we’re close to MVP

  • Atlas
    • FIRST into neon maps and diffuse
    • Literally even neons have separate drawcalls for foreground, dropped items and
    • Then trivially separate into items/background stuffs/foreground stuffs
      • Since we know these will always be drawcalled separately
      • We control rendering, it’s not like unity so we should really be good to go
    • And boom you have tripled the amount of textures we can contain
  • Equip z prawej w jak w ceesie najlepiej
    • equip load pasek bardzo widoczny nad wszystkimi slotami
  • W winter stories czy czymś takim elementem gameplayu jak nie masz ammo jest rzucanie melee i zbieranie z podłogi z powrotem
    • jak w duelach

  • More camera modes for Spectator

  • Multiple official servers
    • Picking best
    • Do this once we have pinging implemented, which we want with the list of community servers
  • Consider a tab for “official” and “community” in browse servers
    • Actually maybe no because it’d be nice to have it condensed and compare best pings quickly

  • “Favorites” tab

  • Quick play
    • From servers that have any players, find the one with best latency

  • Make server save RCON changes to the config

  • Inventory improvements
    • When a melee weapon is involved in akimbo, pressing its corresponding button results in a throw
    • When two melee weapons are wielded, pressing its button results in both thrown at once

  • “Are you sure you want to quit?” from the ingame main menu

  • fix vim’s premature build successful

  • implement a keycombo to dump and open current logs (multiple clients)

  • Nerf PRO90
    • less bullets please
  • Make the bomb a lot lighter
  • Fix bomb planted reward
  • Game Guides

  • Recon grenade

  • We have to renew that certificate or people will be unable to update the game
    • somehow it still works
    • as it’s done over ssl
  • Randomized tips
    • Stay still to regenerate stamina faster.
    • Double-tap R to drop the old mag for a faster reload. Near-empty mags are dropped automatically.
    • Melee blows are stronger when you run.
    • Electric Shield lets you live longer, but slows down your movement.
  • Config folder with delta configs
    • like in linux
    • sorted by filenames

  • It would be nice if we didn’t crash on loading a corrupted map, at least if we manage to hit an ensure before sigsegv

  • Pulse “your own” color for a funny effect

  • Fix owner tracking for grenades as they get deleted and the death summary can’t know whose that grenade really was

  • Sounds for grenade throwing
    • Flashbang out
    • Throwing frag
    • Frag out
    • EMP
    • Interference!

  • maybe don’t save the last launch unless in debug

  • problem: footstep smokes might still be visible when we artificially increase zoom
    • well the client itself knows everything anyways
    • so we can force the zoom to be exact
      • if someone has more than 1920x1080 they technically have the advantage but the footsteps would still be hearable
        • one could still see some distant items thrown but come on
  • ImGui popup for progress on loading atlases
    • the imgui atlas itself will load quickly enough
  • Initial popup about the zoom change
    • or scale automatically to the server’s fog of war size?

  • Consider implementing a radius range for item pickups

  • Shouldn’t we always calc reloading movement to show context-lessly?

  • Force proper drawing order of attachments as this really looks bad on the rocket launcher
    • At least on guns?
  • Lag estimates for other players
    • send a byte for count, and a byte for each player
      • they can be consecutive and we don’t have to include ids
      • the order of existent ids implicitly known
    • can be sent over the solvable stream
      • later we could introduce advantages for players with higher ping, although may not necessarily be good idea
  • Detect when the initial state is being sent to optionally display progress
    • Possibly by returning some info from ReceiveBlockData struct
  • Properly implement bots
    • Now they can break the client logic
  • large block sending augs api
    • asynchronous: deliver_large_data
    • large_data_delivered() const
    • always one large data is sent at a time
  • enum-less variants of enum_boolset and enum_map
    • useful for quick mapping users in an array
  • So that we can change tickrate dynamically, timing logic can, instead of steps, simply hold the floating point (double) values of the seconds passed
    • alternatively hold unsigned long longs?
    • also store last_fixed_delta for easy cosmos::get_fixed_delta access
  • Usability fixes to editor
    • Setting specific names to entities
    • Complex instantiations
      • I struggle to see any that are necessary before DM
        • even manual transfers arent that necessary
      • Guns with magazines already
      • For now expose transfer GUI?
  • enum_boolset begin and end should probably return enums
    • simpler lua i/o
    • no need for for-eaches
    • nice exercise for iterators
  • When joining a game, allow choosing the flavour of the character
    • Each might have completely different properties or animations
    • Effectively implements a class
  • Arbitrary pasting of entities
    • Vastly useful for importing stuff from testbed maps into existing ones
      • Let alone between community maps
    • This is pretty complex.
      • If we allow pasting across workspaces, we might even have to copy actual asset files.
    • Won’t matter until after deathmatch stage
      • Surely?
    • The editor will have to construct the command tree, like
      • paste_flavours + paste_entities, if no requisite flavours are found inside the project at the time of pasting
        • the clipboard will have both the entity and flavour
      • the editor’s clipboard can actually become…
        • paste entity flavour + paste entity command, stored, waiting to be executed!
          • the move itself won’t need to be stored
    • Cut is just copy + delete
  • Allow to change the tickrate in the non-playtesting?
    • This will be a session setting, really
    • We’ll just restart on re-tick

  • Overwrite notice in editor

  • Refactor: conditional log macro

  • Always initialize the hotbar with some values if it is not yet initialized

  • Commands refactor: separation of undo and redo state
    • redoer and undoer objects
      • each has only the required information
      • redoer returns an undoer
      • redo state is always the same
  • Balance colliding bullets so that damage to be dealt is subtracted from the stronger bullet
    • Possibly reduce trace size then
  • Local setup should record session live
    • This is really important in the face of bugs.
    • Or just get rid of test scene setup for now and let it by default launch a scene in editor that records inputs
  • Fix rendering order of dropped gun attachments
    • Also make it possible to always render them under?
  • consider having entity guids in components instead of entity ids for simplicity of network transfers
    • question is, won’t we anyway be sending the whole pool state?
      • it’s just several bytes overhead per entity
    • there ain’t really that many and it will be greatly useful
  • Fill new workspace with test scene essentials
    • This would prevent image ids in common state from being invalid
      • Probably less checks to make, in GUI as well
    • Can make those the first in test scene image enums so that we can stop importing images after some point

  • copy gfx and sfx folders on save as…

  • let particle definitions be split into the invariant and variant parts, like components
    • pro: better cache coherency

  • templatize some dangling memory stream arguments

  • in go to dialog, make selection groups appear as the first
    • later we might just make a variant of several types instead of entity_guid in match vector
  • fix saving of the editor view state when the work is saved to some non-untitled location
    • notice that, we might later introduce some caches for selections to improve performance
      • e.g. only always calculate the selection’s aabb and rotation centers once
    • we might also want to encapsulate setting panning cameras and selections to have stacks
      • where we’ll able to set modification flags easily
    • therefore let’s do it when we get to these stacks.
      • because camera & selection state isn’t all that critical
        • and untitled work for testing is always saved anyway
    • we can just always write the camera to disk?
      • it’s bad though as it incurs disk activity every time
      • and selection state isn’t that small
    • we could either set an “is_modified” flag or track it in history
      • not much modification points for selections, so quite viable

  • pass display size and implement augs::get_display

  • some hint for windows unicode in imgui: https://github.com/ocornut/imgui/issues/1060

  • finish work with atlas and sound regeneration
    • regenerate only seen assets
    • always load diffuse map with the corresponding neon map to avoid unilluminated objects near the camera…
      • …even though many diffuse maps nearby have been loaded
      • perhaps we won’t really need the separation between diffuse and neon, because maximum atlases will be huge anyway

  • Shake on shooting with a sniper rifle

  • Dashing
    • Assigned to space, since we don’t have jumping anyway
    • pure color highlight system could be used to add highlight the dashing entity
    • Gradually increase walking force over time
    • The more the speed during dash, the stronger the dash
  • Note: drone sound (sentience sounds overall) will be calculated exactly as the firing engine sound
    • Sentience humming caches
    • These don’t need their playing pos synchronized
  • Since mouse motions will probably be the bottleneck of network communication, both coords will usually be able to be sent in one byte
    • Range would be 0 - 16
    • There would be a bit flag for when it exceeds the range
      • Which will actually be often and then we might use two bytes
        • mouse motion should never really exceed 255
          • will also make it harder to use aimbots that instantaneously change mouse location