A slime made without finalizeSpawn keeps 20 health where the probe's has 1,
the probe doesn't report a piglin's brain, and a NoAI bee's flower cooldown
runs one tick apart; none of that is the Peaceful despawn rule.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Two changes so `parity.py --changed` keeps picking the right harnesses once code moves into a crate of its own:
- A harness recorded before the move names the old paths. A changed Rust file of a crate the recorded crates
link (their path dependencies, transitively) but don't name now counts, and the reach search decides.
Before, such a file matched no record and selected nothing.
- A library's root opens once reached code names the crate (`horizon_world::`, `pub use horizon_command as
command`), instead of always. Every crate root was open, so any item of a new crate whose name matched a
name used anywhere counted. On main the selection is the same for 15 of 16 sampled files; for
horizon-world/src/nbt.rs one harness that reached it only through a name match drops out.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Every Railway build of the registry has failed since horizon-mods gained the optional horizon-shader-api
dependency ("no matching package named horizon-shader-api"). Built the same context locally: OK.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Rabbits get BreedGoal and TemptGoal with rabbit food. Damage to an animal
resets love. Horses, donkeys and llamas need canParent (not ridden, tame,
adult, full health) for both parents, a sitting wolf is no partner, and the
breeding experience orb needs the mob_drops game rule.
Gift loot (chicken eggs, sniffer seeds) is now rolled with the mob as the
loot context's this entity, and chicken variants are known to loot
predicates, so each chicken lays its own egg colour. A block a player
breaks now has the player as this entity, so snow layers drop snowballs
and chorus flowers drop themselves.
The probe's output is what the harness compares with vanilla; the codec
round trip lives in the unit tests, and its unwrap on a cost whose
components name an unmapped registry entry could abort the probe.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
AbstractHurtingProjectile.tick calls applyInertia first and then moves by
the new delta movement; Vec3.normalize leaves a near-zero vector at zero.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
🤖 Agent: game.raycast, mod API level 2. CI is green on c23b4e8e4. In game, in-frame and on the server thread, the ui guest's /raycast passes all four checks: glass at 2 blocks, nothing at 1, gold after /setblock, the sky is no hit. The ported Orbital Laser aims with it and fires in both modes. Queued.
FireBlock.getIgniteOdds returns 0 for a waterlogged state, so canBurn is
false there and the smoke takes the other branch.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
EntityType.create returns null for a type that isn't allowedInPeaceful once
the difficulty is Peaceful, so the peaceful_* scenarios threw in the oracle.
Their mobs stand for ones already in the world when the difficulty changed.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
🤖 Agent: harness-only change (skip-release). The old harnesses fail under server stalls, the new ones pass; 20 affected harnesses pass locally, and the two changed ones rerun after merging main.
Built on the Client of the harness-await branch (#2114). Under 2.5 s server stalls the old harness fails (a command's
feedback filed under the next step) and the new one passes 3 of 3.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
🤖 Agent: The test scripts for server mods. CI is green on d64bd7c52, and the PR is now based on main after #1833 merged. I ran pr_ready --quick on this head merged with main (b8027d4b1): fmt, clippy, the horizon-mods and welcome tests (48) and the launcher tests (132) all pass. The only affected harness is lan_integration, which runs in CI. Queued.
fireworks.rs keeps main's rocket blast damage next to the client burst;
lan_e2e.py takes main's CHECKS and server-thread default plus 'fireworks'.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The Add/Edit Server, Direct Connection and Friends screens and Create New\nWorld built their titles, field labels, defaults and buttons from English\nliterals, so every non-English language showed English there. Route them\nthrough the current language with vanilla's keys:\n\n- ManageServerScreen: manageServer.add.title/edit.title, enterName, enterIp,\n selectServer.defaultName, gui.done and gui.cancel.\n- DirectJoinServerScreen: manageServer.enterIp (the addServer.enterIp key is\n not in 26.3), selectServer.select and gui.cancel.\n- WorldCreationUiState: selectWorld.newWorld.\n- Select World: selectWorld.title and selectWorld.versionUnknown.\n- DisconnectedScreen: disconnect.lost and gui.toMenu.\n\nA language missing a key keeps the English fallback, so en_us is unchanged.\n\nCo-Authored-By: ZCode Agent <noreply@zcode.ai>
Menu text drawn with characters the bitmap sheets lack (Chinese, Japanese,
Korean) is rasterized into a reserve of the shared atlas at run time, and the
font caches the atlas rect it got. The loading screen hands over to the game
about two seconds after the title shows by building a new renderer, whose
atlas texture is made from HudAtlas.pixels alone: those never hold a
run-time glyph, while the cache still answers with the old rects, so every
such character already drawn went blank. Latin text, straight from the
sheets, stayed. build_renderer and loading_renderer now forget the cache
before the new texture exists, so the characters that are drawn again are
placed into it, as ui_theme clears its own cache when the atlas changes.
Fixes#1824, fixes#2040.
capture(true) now calls reset_inventory_clicks; the fixture compiles the
method body against a stub Game.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
🤖 Agent: verified: pr_ready --quick green, 30/30 affected harnesses pass locally, CI verify green. Internal: one WGSL prefixer for the shader-pack stages (no visible change).
`game.raycast(distance)` returns the first block along the player's view
(fluids pass), up to 1024 blocks: through the loaded chunks as the
crosshair's ray goes, then across the distant terrain's ground heights
(`Lod::ground`, from Stink-O's fork). The game casts it only while a running
mod imports the function. It is API level 2, so older Horizons refuse a mod
that calls it with "update Horizon".
Co-authored-by: Stink-O <180130752+Stink-O@users.noreply.github.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
🤖 Agent: The launcher half of the mod API levels. CI is green on c8464c6bf, and the PR is now based on main after #2080 merged. Launcher tests pass, the new browser test included. The level mark is read from a real fast build of the game. Queued.
The jars an import fetches went through the proxy whatever the host, so a host
that only answers directly failed there while the launcher's own requests to it
went direct. net::agent now takes the URL and applies the same NO_PROXY and
loopback rules the launcher's own traffic gets.
Co-Authored-By: WorkBuddy AI (DeepSeek-V4.1-Flash) <noreply@workbuddy.cn>
The world side checks Class::BaseFire; the tag is a data-pack list that can
differ from the class.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Dragon fireballs were a server-side list like the breath clouds, so a server's players never saw them. Each is now tracked as a minecraft:dragon_fireball; the client keeps it in the same list, flies it on as AbstractHurtingProjectile.tick does, snaps it to the server's moves and draws it with its dragon_breath trail.
🤖 Agent: Mod API levels. CI is green on b12c89650. Locally: horizon-mods lib 31, api_reference 4 and sandbox 26 pass. A dummy component that calls a function from a newer level is refused with "update Horizon". Level 1 is render.set-minimap and the players.* functions; level 0 matches 0.2.67 to 0.2.69. Queued.
The breath clouds are kept as a list, not entities, so the server thread drew their particles into its own copy and players never saw them. Each cloud is now tracked as an area_effect_cloud (radius, waiting, particle), and a server's client draws it as AreaEffectCloud.clientTick does.
Applying packs dropped the window surface and created a new one for the renderer built on the same device. On Wayland with Mesa (RADV) the new surface came back lost (ERROR_SURFACE_LOST_KHR), the new renderer failed, and restoring the old one failed too, which crashed the game.
🤖 Agent: Merged after the contrib-gate-16 review: checked against 26.3, the in-game checks and lan_e2e in both modes, and the 79 affected harnesses on the integration tree (see the gate comment). Thanks!
🤖 Agent: Merged after the contrib-gate-16 review: checked against 26.3, the in-game checks and lan_e2e in both modes, and the 79 affected harnesses on the integration tree (see the gate comment). Thanks!
GitHub started both 10-26 schedules over an hour late, so the hour check skipped both and no audit ran.
lan_integration takes ~2,400 s under load, past the 1,800 s per-harness limit.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
🤖 Agent: gate verified on e3e08911b (main merged). fmt, clippy -D warnings and horizon-client tests pass. lan_e2e with the server-thread host: 131/133, and cursor/falling pass on rerun in both modes. 14 in-frame checks pass. agent_control_e2e: only manage fails, and it fails the same way on main. 82 of the 83 affected harnesses pass. lan_integration flakes under load and is now named-only (nightly). Queued.