[pull] main from fishfolk:main#2
Merged
Merged
Conversation
Install the bones network debug egui plugin / add a button to open it from debug window. Companion PR to fishfolk/bones#345
Add a steam_package.sh script that pulls releases from github, rezips as .zip such that they can be uploaded to steamworks for respective platforms. For linux, we also install some launch scripts. There's logic here to fix #904. Isn't pretty but seems to work, scaling is correct in steam deck game mode + desktop mode. Should not impact other linux platforms. Need to test again with steam deck with external display, but I think this should work based on last I was exploring solutions to #904.
companion PR for: fishfolk/bones#346 frame time window re-added to bones, this enables viewing it.
…pdate velocities to keep behavior (#925) **Overview:** Delta time in physics update previously had a scale of 60 due to a prior fix regarding timestep. This change removes the scaling of physics timestep by 60, and adjusts velocities / forces (gravity) to account for this so behavior does not change. **Impact:** Will help with ragdoll work I am doing, velocity metadata can be used with objects simulated in rapier and on our kinematics without having to get messy scaling them back and forth, or dealing with a very large timestep in rapier (which did not go well in my experiments). units of velocity and gravity now relate more intuitively to position + sizes of objects in terms of setting velocity based on how far a contributor may want something to move per second. **Changes:** - velocities all scaled up by 60, gravity scaled up by 3600 (60*60) as this is multiplied by time twice. - Gravity values are now quite large, unless we want to change unit of gravity and add back in a magic scale factor will have to roll with this I guess. I think it's ok this way, new contributors may reference global gravity metadata to get an idea of what to set a specific body's gravity to. - apply_rotation now uses delta time (kinematic rotations should now be consistent regardless of frame rate) - Fix bullet position update from velocity not being scaled by delta time - Fix setting velocity.y = 0 when grounded comparing against global gravity (now corrected to use body's gravity) - Fix sproinger velocity condition including gravity incorrectly (don't think we need gravity in this at all) **Other Notes:** I am hoping I caught all of the values that needed updating. Testing movement / jump, throwing items, sproinger, bomb bounces, it looks right to me. - One thing I noticed is that some item metadata includes a "angular_velocity" value which is used as spin factor for item throw. Spin scales based on the throw velocity, so these spin values do not need to be updated. This is more of a spin factor than a true angular velocity, may consider renaming in future. - There are a few values hard coded that we could pull into metadata, but I didn't want to make footprint of change any larger than necessary. - some metadata for player movement is duplicated between player skin metadata, may want to pull this out to single metadata location in future.
This is the base of core changes to support ragdolls, allows us to simulate dynamic bodies in rapier. DynamicBody component may be added to an entity that has KinematicBody, and if enabled, body will simulate physics. Transitioning between kinematic / dynamic and adjusting rapier data + collider is handled on next physics update. This change should have no impact on existing kinematics, scene queries, etc. Usage of these changes will come in other PR soon. **Switched to fork of rapier (temporarily hopefully):** I switched us to a fork of rapier to fix a bug in which changing a rapier rigidbody from kinematic to dynamic will not put it in active dynamic set, so gravity is not applied. It seems they rely on contact graph to add body to this set, but if not in contact with anything it breaks. I have a PR open for a fix in rapier. - I also updated version, slight change in debug draw code is because of changes there. **Some notes on changes:** - Some helper functions are included for pushing simulation commands for dynamic body that apply changes to rapier body before next physics step. - There is a mutate_rigidbody function on CollisionWorld that makes accessing rapier body for immediate mutation slightly more convenient. - Previously bodies were created as dynamic (but we used them as kinematic anyway), now they are actually flagged as kinematic in rapier. This does not impact collision events at all. There's a decent chunk of physics stuff in collision module now, at some point maybe should detangle this, but probably fine for now. I renamed some functions to make things more clear.
…simulated bodies (#928) Another chunk of features implemented to support ragdolls + dynamic bodies in Jumpy. ### **Contact filtering:** Add bit flags for rapier `CollisionGroup` and `SolverGroup`. - The Collision group filters both contact forces and events, these are configured such that there is no change here in collision events. - Solver group then filters contact forces, this is used to make sure dynamic actors only simulate collision with tiles (solid + jump through) and solids, not other kinematic (or dynamic) actors. ### **Jump Through for Dynamics:** Dynamics collide with jump through tiles. A contact modification physics hook is implemented so dynamics moving upward through jump throughs have contact thrown out, and kept when falling. - Only the contact filtering `PhysicsHook` is used, and is only enable on dynamic bodies while they are simulating. ### **Changing actor shape:** I found that using a capsule works better for player ragdoll then rectangle, which requires being able to change the shape of a collider. A helper function is added to do this correctly, changing shape is only currently supported for actors. ### **Handling `Stop` collision events for removed colliders:** Changing an actor's shape involves removing collider and re-adding it. This triggers a stop event from rapier. These events provide collider handles, yet we map this to entities from userdata. The consequence of this is that when the stop event is dispatched, the collider handle is already invalid (generation incremented), meaning we no longer can access user data to determine what entity should be removed. - We now have a function that should be called on removal of collider in `CollisionCache` that temporarily caches the handle + entity, which is used as fallback when receiving stop event if collider handle is not valid. This is not a great system, if someone removes a collider and does not call this, our collision event will not be removed causing bugs. Might be able to make this better, but for now will just have to make sure if we get fancy removing a collider (which is already a delicate / involved operation) this is called. The function to change actor shape calls this for us.
### Ragdoll State Player may toggle ragdoll state with `F` or `Right bumper`. - player body switches to capsule and simulates physics in ragdoll state. - movement during ragdoll twitches the body slightly, small velocity + animation change. ### Minor Related Fixes - set position with respect to parent no longer forced to 0 for all objects. Used this to tweak capsule position. - Attached entities now rotate with parent. ### Animation This will need another pass, for now I re-used some stuff we have. Making the corpse more obviously dead would be good (though I guess being able to reasonably play dead may also be amusing). https://github.com/fishfolk/jumpy/assets/35712032/9bdcbc0c-146b-4339-a7bd-4ed1ca1bb44d https://github.com/fishfolk/jumpy/assets/35712032/f9004638-fc03-4ee6-8e6f-0f7f0ab1f3a3 https://github.com/fishfolk/jumpy/assets/35712032/5407bcf3-e9c9-4aa3-83e1-ba95adc3fcac
Closes #930 - Adds a `PlayerCommand` to drop a player's hat and reset its properties - Replaces all instances of manually dropping a hat to use the new convenience command
…from metadata (#935) I opened a PR in bones to expose these settings for networking, instead of hard coding in engine. (fishfolk/bones#354) The defaults of these values were also changed in bones such that things are feeling much better in network play with ping around 100ms-150ms. Now we may tune in metadata, though settings only take effect if new network match is created. This will not pass CI until bones PR is merged + bones version updated.
…n with dynamics (#931) define some collision info for jump through wood tiles, used to create accurate colliders for collision with ragdoll / dynamics. The standard tile is still there for all intents and purposes. If there is an extra collider defined, this collider is also added, and is set to solver group such that dynamics will collider with it. - This additional collider is not used for collision events. - if additional collider is present, original tile collider no longer has solver group for collisions. This fixes dynamics colliding with the invisible chunk on thin tiles such as jump through wood. I think we're done with the trickle of auxiliary features and next PR should be ragdolls 😆
Fixing a crash in release that was in 0.11.0
…#946) Should switch off git revision once new rapier release is published.
The new blunderbuss is done, this closes #579. I used the artwork from #152. `bullet_spread` and `bullet_count` are open for changes on `assets/elements/item/buss/buss.yaml`. I also made a quick artwork as the musket bullets explosions are too big for the buss bullets size, but it's basically a smaller musket explosion. My idea was a mid-range shotgun but it could be easily modified through `yaml`. Blunderbuss will spawn in level 1, 3, 4, 7 and 11. https://github.com/fishfolk/jumpy/assets/107818001/93ee26ef-3e94-49a2-8bce-e7484eb1ed0d
Periscope ported, this closes #151. It's similar to the musket and is mainly useful if the enemy is higher than you. Used the artwork from #151 but the concept art confused me a bit as it says to be "similar to the musket" but in the art it looks like more a shotgun as it has some bullet spread, but adding a very low bullet spread would be a nice idea here so let me know if you think it should be used. As it's very similar to the musket, it had a certain amount of code duplication, which I think we could discuss together and come to a decision whether to generalize something elsewhere or whether we want to make adjustments to add more uniqueness to the designs.  
Closes #951 - De-duplicates dependencies: - `async-lock`, `base64`, `getrandom`, `linux-raw-sys`, `parking_lot`, `parking_lot_core`, `rand`, `rand_chacha`, `rand_core`, `rustix`, `synstructure`, `wasi` - Duplicates dependencies: - `erased-serde`, `polling` - Fixes security advisory due to `failure` crate removed from dependency tree - All CI tasks now pass! 🎉 --- <details> <summary>Compare duplicated dependencies</summary> <pre> function get_dups() { cargo deny check 2>&1 | sed -En "s/^warning\\[duplicate].*('.*')/\1/p"; } <br> git switch main get_dups > before.txt <br> git switch <ref-you-want-to-compare-to-main> get_dups > after.txt <br> git diff --no-index before.txt after.txt </pre> </details>
This PR closes #573. I added `bullet_spread` system to the machine gun because I think it balances the game and looks good, and added `empty_cooldown` because it would be really boring if `gun_empty.ogg` kept playing at the same frequency as the normal cooldown, so in this gun we have the shooting cooldown and the empty sound cooldown. I also made a very slight change in the main system for using items because it had to be able to shoot constantly, so now you can keep shoot input pressed with any weapon. Something I think is important to mention is that the shooting animation is synchronized to the rate of fire, so if you change the rate of fire the animation will also change. It could be easily changed with a new variable in the .yaml so let me know whether you think it should be faster or slower. The machine gun has two states (idle and shooting) and its only function is to reset and start animations.
I did what @MaxCWhitehead told me earlier, created a `bullet_spawn_offset` field in `.yaml`. I added this to all the weapons because it makes it easier for new contributors to locate and modify in case we change the artworks.
…in QBVH (#960) I've been frequently hitting a crash / bug in parry2d QBVH involving incremental update. Sometimes it crashes, sometimes is a framerate death spiral / fully locks up. ``` thread 'main' panicked at /Users/max/.cargo/registry/src/index.crates.io-6f17d22bba15001f/parry2d-0.13.6/src/partitioning/qbvh/update.rs:367:54: attempt to add with overflow ``` Appears to be been fixed at dimforge/parry@7df2728 in parry2d 0.13.7. Updated rapier to use a fork with parry2d update as not updated in rapier yet.
…every frame (#962) See #961 for context. There's a bug in parry2d involving incremental update failing to balance broadphase BVH. Fully rebuilding BVH avoids this issue, and we can get away with it due to few bodies in our game. This was our previous behavior before supporting dynamic bodies and relying on incremental update in physics pipeline, which turns out to be broken.
…#966) Update bones for the following fixes: fishfolk/bones@6b8d743 (fix: Fix race condition on if EguiCtx is initialized before bones game) fishfolk/bones@cbe537f (fix: Fix Zero'd DenseMoveDir not representing no input.)
… max prediction (#967) Update bones to get a fix for issue causing desync in 3+ player matches.
…1017) Use AudioCenter from bones now that it has been moved there, and remove from Jumpy. Added a trait extension to continue playing music from kira `StaticSoundSettings`, to avoid having to put all of the params in `play_music_advanced`, as the looping is configured a bit differently. The addition of `force_restart` in PlayMusic event has been helpful, now music does not restart each time switch menu pages. Resolves #1016
fixes https://github.com/fishfolk/jumpy/security/dependabot/18 (we should probably update the other actions too)
We updated download-artifacts to v4 but not upload, seems to have broken release pipeline, hopefully this fixes.
Add support for some environment variables, only available in debug builds, that allow for quickly loading into a single-player game. ## Examples To launch straight into a single-player match with Fishy, the default player, on Keyboard1: ```shell TEST_MAP='Dev Level 1' cargo run ``` You can also specify another player: ```shell TEST_MAP='Dev Level 1' TEST_PLAYER='Sharky' cargo run ``` And even provide a hat: ```shell TEST_MAP='Dev Level 1' TEST_PLAYER='Sharky' TEST_HAT='Cowboy Hat' cargo run ``` Or, auto-select a player, launching directly to the map select menu (again, optionally with a hat): ```shell TEST_PLAYER='Orcy' TEST_HAT='Unicorn Hat' cargo run ``` All of these examples set Keyboard1 as the control source for the player. You can provide a different source like so: ```shell TEST_CONTROLLER='Gamepad' TEST_MAP='Dev Level 1' cargo run ``` Any combination of one or more of these four environment variables will skip straight to either the map select menu or directly into a match if a map is provided. Fishy with no hat on keyboard1 is the default if only a map is given.
…te (#1032) This updates bones and uses new matchmaking flow. Includes using the new matchmaker rng seed, and gilrs update. Some other little fixes included: - "Waiting for players" text now displays again when matchamaking - If start matchmaking online and switch to lan tab, matchmaking is now cancelled. Gilrs update confirmed fixes: #989
Fixes #933 Disabling only gamepad input seems to make the input behave as one would expect. Arrow keys, backspace, etc work like normal, and bound keys (e.g. A, D) only update the text and no longer move the caret. Unfortunately I don't have a controller to test with, but I assume this will work with those kind of inputs too. --------- Co-authored-by: Max Whitehead <max.blackllama.alt3@gmail.com>
Follow-up to #1023, partially reverting it as there is a simpler fix, and fixing the remaining bugs.
Fixes #963 ## Changes - [6781542](6781542) `PlayerSlot` is now (sort of) a state machine, starting out `Empty` then progressing through `SelectingPlayer` and `SelectingHat` until becoming `Ready` - Slot for a network player only uses the `SelectingHat` and `Ready` states for simplicity - Angled brackets `<` and `>` show around the thing being selected (player/hat), brackets now don't show when ready - [a2393c4](a2393c4) fix bug where label has incorrect text for 1 frame when pressing Menu Back to remove player from slot - [f636d33](f636d33) Upgrade bones to latest - Set gitattributes for `*.ftl` files so that they are treated as text for diffs and not binary files - Ignore advisory `RUSTSEC-2024-0370` proc-macro-error is unmaintained (see [related bones issue](fishfolk/bones#479) and [the advisory](https://rustsec.org/advisories/RUSTSEC-2024-0370)) This also fixes a bug I was seeing on `main` where another player in a LAN lobby pressing A/D would behave as if everyone pressed A/D, e.g. pressing D to select the next skin would change to the next skin in both slots. **Rationale for the rough state machine:** Previously there were 3 states for a player slot: Empty, SelectingPlayer, and Ready + SelectingHat. This state was maintained by two booleans, `active` and `confirmed`, which tell you whether the slot is not empty and whether the player has readied-up respectively. Together they represent the 3 states as: (`active/confirmed`) `false/false`, `true/false` and `true/true`; with `false/true` being an invalid state. Now that there are four states to the ready-up process, one way of implementing this is to add a third boolean, but this introduces 3 more invalid states. Let's say the boolean is called `selectedPlayer`, these are the possible boolean combinations and the states they represent (`active/selectedPlayer/confirmed`): ``` false/false/false Empty false/false/true <invalid> false/true/false <invalid> false/true/true <invalid> true/false/false SelectingPlayer true/false/true <invalid> true/true/false SelectingHat true/true/true Ready ``` The implementation I opted for is a simple state machine. While it added a lot more code to some areas of this module I found it made reasoning about some sections much easier. It also made explicit certain edge cases in the `handle_match_setup_messages` system where e.g the client could theoretically receive a confirmation message from a peer when it had not yet received any player selection messages. A warning log is now generated in the unlikely case this occurs.
…ready (#1040) Fixes #1036 and couple other things along the way Not entirely sure what caused this - but I think I introduced it when updating for new matchmaker interface in bones. We would allow online matches to continue if >1 players are ready and no unconfirmed players. The issue is that in online remote players who had not made a selection were in `Empty` state, and not tracked as unconfirmed. Now for any local + required players, I initialize them to new PlayerSlot type `SelectingLocalControlSource`. `Empty` may be used for online slots that are unused (the last 2 slots in a 2 player game) in online, or for lan, optional slots that may be joined by additional players/AI, but are not required. And remote required players are set to `SelectingPlayer` with default fish + no hat - as next message expected from them is change in player, and they don't need to select control source, we know they are remote. This seems to have fixed the match starting without all players ready, as all required players are initialized to something other than `Empty`. ### Other fixes / changes #### Remembering Hats: I noticed that if you select a hat and then deselect it going back to fish select - the other online players would still see this hat, but local player resets to None. If player selected hat, went back to fish select, then readied up with no hat, the remote players would see hat, but not the local player. This could cause desync. Now When you go back to fish select, your hat is remembered. #### Always display default fish: We were no longer displaying default fish in online when player had not yet selected a control source/fish. Now regardless of if have actually selected a fish, the default fish + no hat is shown (If the slot is a required player that we are sure will select one eventually / must ready up). In local for unjoined slots, no default fish is shown until joined or AI is set. #### Show when remote players ready In online when a remote player is ready, this is now displayed. @nelson137 feel free to review (if you want) as I know you were just in here. Player select sure is complicated lol, but feeling good about the changes I think we're making it more manageable 😅
Fixes #990. Reorder the spawners in the level asset so that players 1 & 2 spawn on opposite sides of the map.
Fixes #1014 (for now). It sounds like a proper fix entails significant work; but to fix the immediate issue we can move the rope tiles away from any others so that those adjacent to them are blank. ## Before <img width="256" alt="before-asset" src="https://github.com/user-attachments/assets/41e8ffb8-4034-492b-9f17-ce182656679b"> <img width="256" alt="before-game" src="https://github.com/user-attachments/assets/e30bf931-ca2a-46a7-86b8-c2e6971e4155"> ## After <img width="256" alt="after-asset" src="https://github.com/user-attachments/assets/86342a5c-cab9-4caf-948b-ce01f68393b0"> <img width="256" alt="after-game" src="https://github.com/user-attachments/assets/df8c7c46-8a4d-4828-8370-fad89c108bdb">
Follow-up to #1041. ## Changes - Fix the rest of the sprite artifacts in `ship_decorations` ## Things Worth Mentioning - Changed levels 3 & 8 to use YAML flow style so that I could find tiles by position (e.g. by searching for `pos: [13, 10]`: Before: ```yaml tiles: - pos: - 1 - 2 idx: 3 collision: Solid ``` After: ```yaml tiles: - { pos: [1, 2], idx: 3, collision: Solid } ``` - Add script to get the layer names in any level files that use specific tile indexes of a tilemap: ```yaml # Find all crow's nests (indexes 30-32) from ship_decorations $ ./scripts/find-atlas-tiles.sh ship_decorations.atlas.yaml '[30,31,32]' ./assets/map/levels/level_3.map.yaml - foreground ./assets/map/levels/level_8.map.yaml - foreground # This tells us levels 3 and 8 have crow's nests in their "foreground" layer ``` - Fix git attributes for text files in `assets/` & `old_assets/` which shouldn't be treated as binary (yaml, txt, ftl)
Added a glug sound when player dies to give additional audio feedback. Here is the audio used: https://freesound.org/people/brittmosel/sounds/529300/ This resolves #986.
…1049) Update plugins for builder pattern, from changes in bones: fishfolk/bones#489
The forked version wasn't working for me when building in a sandbox, so I've bumped it to the upstream one that contained the fix. Signed-off-by: Marcin Serwin <marcin@serwin.dev>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
10 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
See Commits and Changes for more details.
Created by
pull[bot]
Can you help keep this open source service alive? 💖 Please sponsor : )