🤖 Agent: pr_ready passed (horizon-world and client tests, affected harnesses), lan_e2e passed in-frame and on the server thread. GeneratorData::load on the ThinkPad 0.64 → 0.45 s warm. Queued for merge.
GeneratorData::load read everything on the server thread while the chunk
workers waited: the block tables, the structure templates and the block
behaviours now load on threads of their own, and BlockBehavior::load reads
its cached-support table beside the other two. The behaviour exports
(gameplay-facts.json is 55 MB) parse from bytes in memory instead of
through a reader.
ThinkPad (2 cores), 3 alternated runs: 0.63 -> 0.45 s with the behaviour
snapshot cached, 2.40 -> 2.07 s on a first start.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
🤖 Agent: Thanks vlesmentie! Matches BigDripleafBlock.entityInside and canEntityTilt (the 0.6875 height check): a big dripleaf tilts under a player standing on it. Tested with the wave 2 batch: fmt, clippy and the horizon-client and horizon-world tests pass, and all 127 affected harnesses pass. The chunk_generation and decorated_region reruns took 8 minutes with no chunk differing. lan_e2e contacts and the rest of wave 2's checks pass with the host in-frame and on its server thread.
🤖 Agent: requeued after a runner-setup parity failure (subtitle: missing sounds.json on the runner); 30/31 harnesses passed, the 56 selected passed locally. Queued by perf-next-2.
Both branches changed how the harness finds the mod jars: resolve every jar
through jar_path (jars, then pack-jars) and keep the profile packs check.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
🤖 Agent: merged after verify (bugs-dezerx-1's fix for a player's bug list, self-checked with pr_ready); harnesses run on main by parity-main, which gates releases.
The snapshot's key hashes the 70 MB of source tables it was resolved from.
The client's tables and the world generator's are loaded from the same
folder, so a world opening hashed them a second time: a third of a second
of the server thread's start on a 4-thread laptop, before any chunk. The
hash is now kept for the process, by each file's length and modification
time; a changed file is read again.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The 128 stronghold biome searches (two seconds of CPU on a 4-thread laptop)
ran when the first chunk of a new world looked up its structure starts, so
nothing near spawn was generated until they were done. Each search ends
within 8 chunks of its ring chunk, which is known before searching, and
strongholds lie 1,400 blocks out or more: a chunk with no ring search in
reach (exclusion zones included) is now decided without them, which is what
they would say. They run when a chunk near a ring, /locate or an eye of
ender needs them (ring_positions now searches on demand instead of returning
nothing before a lookup did).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
On the rows along a box's bottom, top and sides every cell is outer, so in a
section no block reaches out of, the rest of the row up to the chunk's edge
is passed in one step instead of cell by cell. The candidate budget still
ends the scan where it did.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The leaf's tilt ticks, its reset and its redstone reset were already in
horizon-world (big_dripleaf_tick, big_dripleaf_reset_tilt), but nothing
started them: BigDripleafBlock.entityInside was missing. An entity on
the ground whose feet are above the leaf's y + 0.6875 (canEntityTilt)
tilts a flat, unpowered leaf UNSTABLE and schedules its tick in 10.
The block tick's inside scan now also passes the host's and the LAN
guests' feet and on-ground flag to Behaviours::entity_on_big_dripleaf.
Fixes#1683.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
BlockCollisions keeps a block on the cells around the box (Cursor3D types 1
and 2) only if it has a large collision shape or is a moving piston. Most
sections hold neither in their palette, so the scan now asks the provider
once per section and passes those cells without reading them: a mob's box
reads about a fifth of the cells it did. Same hits, same order.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
CopperGolemStatueBlock.useItemOn: anything but an axe, an empty hand too,
cycles copper_golem_pose with its sound and block_change game event; honeycomb
on a weathering statue passes on to waxing. Axe scraping, unwaxing and
stripping and honeycomb waxing now count the item as used, and waxing sends
its block_change game event, as vanilla's useOn does. The use_item_on harness
gains statue cases.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Each chunk's compiled rule is specialised once for every band of y that
its vertical gradients bound, and for blocks below and at or above the
preliminary surface when the rule asks for it. Conditions those facts
decide are folded away, so a deep stone block evaluates only the rules
that can apply there (ore veins, the underground rules) instead of
walking the whole surface tree. Conditions only read the position and
the world, so the blocks are the same.
Surface stage, one thread, seed 1234: 2.45 -> 2.08 ms per chunk.
From spec-yungsapi rev 3 §3, §3.1 and §3.2 (impl-questions IQ-2, IQ-3):
- the start piece does not count toward max_count;
- is_priority entries are tried first, in the pool's shuffled order, without
a weighted draw; min_required_depth skips a pick (no draw) while the parent
piece is shallower; ignore_bounds places without the free-space test (its
box is still taken, IQ-12);
- the fallback was already shuffled only when reached.
The port no longer counts these three fields as lost. The layout harness gets
yung-probe and yung-probe2: the spec team's probe pack (built from the oracle
profile) carried as a mod, compared start by start with the oracle's census:
384 of 384 and 32 of 32 starts match.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The twenty unwrap_or_else(|e| e.into_inner()) calls become one lock()
helper, so the bots read as they did with the same poison tolerance.
bound_property moves from worldgen.rs (between DecodeError's impls) to
block_predicate.rs, next to PredicateBlockResolver, whose
predicate_property it implements.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The previous tests of litho.rs are deleted unread (clean-room ruling of
2026-10-25). The new ones build their own block registry from a facts
table and run each processor through process_list against a one-block
world: block_swap's property copy (CTOV's beetroots), references in
order and discard_input, the probe pack's nested condition with else,
apply_random per piece and per block, set_block's providers and its
block entity modifier.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The mod world now takes each saved start whose pieces are all jigsaw pieces
as the oracle saved it (chunk_dump's HORIZON_SAVED_STARTS, through a new
GeneratorWorker::use_saved_start and PoolPiece::from_saved), so the blocks
are checked whatever the layout: Lithostitched lays jigsaws out its own way
(spec-lithostitched OPEN-L5) and the layout harness checks layouts apart.
Lithostitched's wrapper elements place their delegate; 1.21.1 block states
are read in 26.3's shape.
Counts are split between positions inside a piece's box and around them
(terrain adaptation, decoration), and every mismatch class is written to
.local/structure-layout/<profile>/blocks-report.json.
A CTOV profile carries Lithostitched and ChoiceTheorem's Overhauled Village
from a pack's jar folder (.local/structure-oracle/pack-jars).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Clean-room implementation from spec-lithostitched.md §3-§5 and spec-ctov.md §4.
The game reads Lithostitched's processors as its data writes them
(processor/litho.rs): block_swap (only the properties the new block can
have: ripe wheat becomes ripe carrots, beetroot keeps its default age),
reference, apply_random, condition (with else, one processor or a list),
set_block (simple and weighted providers, preserve_state, block entity
modifier) and discard_input; the conditions all_of, any_of, not,
random_chance, true and matching_blocks (input or location, properties).
Their random follows random_mode per block or per piece; Lithostitched's own
seeding is open (OPEN-L6), so a positional random of the world seed and the
processor stands in (impl-questions IQ-7).
horizon-port-pack's new lithostitched module carries the processors as they
are, with the processors nested in them carried, drops those the game cannot
read (lists named by tag, other providers) as Java processors, turns CTOV's
modular_compat into a reference when its mod is ported along (else nothing),
maps lithostitched:jigsaw to vanilla's jigsaw (vanilla layouts, the decision
on OPEN-L5), unwraps limited/guaranteed/delegating elements to their
delegate (limits counted as lost until the next PR) and writes
add_template_pool_elements modifiers as horizon:add_to pools when their
NeoForge load conditions pass.
Deleted unread: legacy_structure_mods/lithostitched.rs and horizon-world's
structure/processor/lithostitched.rs. Blind edits to the legacy folder,
located by exact text or the compiler's error position: its module line
removed, dead_code allowed on ctov (its modular compat is unused now) and on a
test helper, and ctov's modular compat test ignored.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Clean-room (spec-yungsapi.md §2, §3, §5; no mod source read; the legacy
files were deleted unread). horizon-port-pack's new `yung` module carries
`yungsapi:yung_jigsaw` structures and `yungsapi:yung_single_element` pool
elements as the mods write them, so players' packs reach the YUNG draws of
the jigsaw assembler. Fields whose behaviour the specification leaves open
(adapt_noise, max_y/min_y, enhanced terrain, element conditions, modifiers,
priorities, depths, ignore_bounds) are counted as lost by name. YUNG's
other structure, placement and processor types go to the general code
(unsupported or dropped) until the specification covers them.
Deleted: legacy_structure_mods/yung.rs, horizon-world's structure/yung.rs,
the `horizon:yungsapi` keys and `horizon:disables`, which nothing writes
now. Two test initializers of the removed field (enhanced_terrain.rs,
piece_limited.rs) were removed by line from the compiler's error positions.
The layout harness now carries the mods through `horizon-port-pack carry`:
82 of 84 Better Dungeons starts match the oracle piece for piece, as before.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Clean-room, from spec-yungsapi.md §2.1 and §3.1 (no mod source read).
A `yungsapi:yung_jigsaw` structure lays out with the jigsaw assembler and
YUNG's draws: the start pool's file entries shuffled, then the rotation,
then a weighted pick in the shuffled order; each jigsaw's target pool is
drawn by weight from its shuffled entries without replacement, an element
whose `name` already has `max_count` pieces skipped at no other draw.
`yungsapi:yung_single_element` places what a single element places; its
pool entry carries the name and limit.
tools/test_structure_layout_parity.py --profile betterdungeons reads the
mod's own data from its jar into an overlay of the 26.3 resources and
compares the mod world's pieces: 82 of 84 Better Dungeons starts match
piece for piece (60 small, 11 skeleton, 11 zombie dungeons, up to 790
pieces); the other 2 sit in 26.3's sulfur_caves, which 1.21.1 lacks.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The clean-room structure oracle saved, for 31 starts of vanilla structures
(villages, outposts, trial chambers, shipwrecks, ruined portals, desert
pyramids over seeds 1234, 2 and 42), the no-mod world's pieces. A new
structure_pieces probe lists the starts a chunk owns, piece by piece, and
tools/test_structure_layout_parity.py compares them with those saves.
22 of the 31 match piece for piece (every village and outpost that 26.3
still places there, and the horizontal boxes of the code-built pieces).
The other 9 differ between 1.21.1 and 26.3 themselves: 26.3 vanilla's own
pieces there (VanillaChunkGenerationOracle scan) equal Horizon's for all 18
starts that differ from 1.21.1. No vanilla jigsaw fix is needed.
Templates now keep the id they were loaded under, so a piece names its
element.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
predicate_property was the same fifteen lines in block_feature and
state_provider; it is worldgen::bound_property now. The mojang link helpers are
unix-only in use but worth testing everywhere, so they are allowed to look dead
elsewhere, which clears the launcher's -D warnings red on Windows. The bots
harness recovers from a poisoned lock instead of panicking.
Stable Clippy (1.99, what CI runs) does not flag q < -2. || q > 2., so the
helper and its test only served Rust 1.90's Clippy. Fire burnout is not
'intentionally disabled': vanilla's FireBlock.tick reads the
INCREASED_FIRE_BURNOUT environment attribute, set by eight overworld biomes,
which Horizon does not read yet; the comment now says so.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
horizon-client, horizon-world and horizon-protocol each had an example
named `probe`. Cargo copies every crate's examples into one
<profile>/examples/ directory, so parallel parity harnesses sharing a
target dir could run another crate's probe (CI parity run 40055656303:
command_parse ran the horizon-world dispatcher). The examples are now
client_probe, world_probe and protocol_probe ([[example]] in each
Cargo.toml; the files stay where they are), and every harness builds
and runs its own crate's name.
tools/parity.py recognises the new names when it counts the probes a
harness ran, and renames the old `probe` in recorded artifacts to the
crate's target (from the record's -p, else its probe sources), so the
prebuild still builds them instead of dropping them as unknown.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
InteractWithDoor was a stub, so a villager whose path went through a
closed oak door walked into it and stayed there. In generated villages,
villagers who were inside a house never reached a job site, a bed or
the meeting point.
Like vanilla, a villager now opens the closed mob-interactable doors on
its next path step (DoorBlock.setOpen: both halves, the door sound and
the block_open game event), remembers the door it leaves and the doors
it opened, and closes them once it is more than one step past them,
unless another villager within 2 blocks is walking through. A villager
going to sleep closes them too.
The mob events apply after the mob loop, so the door opens one tick
later than in vanilla, where DoorBlock.setOpen changes the world during
the villager's brain tick.
lan_e2e's new doors check shuts a villager in a room with a closed
door and a free composter outside. Before, it never opened the door;
now it opens it, becomes a farmer, and the door is closed again on the
server and on the guest.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
#1432 deleted both as unused; #1424 (level events) and #1425 (scoreboard)
merged after it and call them, so horizon-client's binary failed with
E0599 at level_particles.rs:54 and scoreboard.rs:338. Both functions come
back unchanged from 4d7f30f14^1.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
datafix: LeavesFix's getSideMask is ChunkPalettedStorageFix's, line for
line; chunk_leaves uses chunk_flatten's. mobs.rs: push_tests takes
sensor_scan_tests' 46-line cow MobType instead of its own copy.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
main.rs now only lists the probes; dispatch.rs runs them. The runner leaves
main.rs out of a probe-binary harness's dependencies and keeps dispatch.rs,
so adding a probe no longer reruns (or flake-checks) every harness of its
crate. It also follows crate::<probe> references between probes.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Eight benches (block_facts, chunk, entity_geometry, json_readers, load,
save, streaming, tick_chunk_membership), lan_flat, save_parity and
vanilla_diff: no harness, tool, doc or workflow names them (review plan,
section 1, dead code). -2,222 lines; Git history keeps them.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Now that the probes are modules of one binary, the helpers copied between
them live once in examples/probe/support: the collision probes' shape
printer (emit, values, box_values, bounds, 8 copies each), the canonical
NaN bits of f64 and f32 (50 copies), the registry JSON-to-NBT reader (15)
and the JSON block position (4). -1,097 lines of copies.
f64::NAN.to_bits() is 0x7ff8000000000000, the literal some copies used,
so every probe prints the same bytes.
tools/parity.py follows include!: three probes include another probe's
source, and stay dependent on it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The 168 horizon-world examples the parity harnesses run become modules of
one example, examples/probe, that dispatches on its first argument;
crate::args() gives each the command line its example saw. Linking ~170
binaries, each instantiating horizon-world's generics again, was most of
the parity job's probe build (review plan, section 1).
The harnesses build `--example probe` and run `probe <name> ...`.
tools/parity.py counts only the probes a harness ran, and the support
modules they name, as its dependencies, so editing one probe still reruns
only its own harnesses (and PARITY_RUST_SCOPE keeps working).
The six manual tools (benches, loot_fill, structure_blocks, vanilla_diff,
attribute_memory, behaviour_load_profile) stay separate examples.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
93 pub items (78 functions and methods, 14 constants, 1 struct) in horizon-client,
horizon-world, horizon-core, horizon-friends and horizon-server-files had
no caller anywhere in the repository: not in the crate, the other crates,
the examples, the tests or the binaries. rustc can't flag them because
horizon-client and the others are libraries, where a pub item counts as
used. The imports and the private helpers only they used go with them
(title.rs date_string, goals.rs flag_index).
Also drops the #[allow(dead_code)]/#[allow(unused)] items that only hid
dead code: chat_debug_fill, merchant_title and Goals::locked_by had no
caller, and four stubs (_keep, _unused, _rule_kind, const _) existed only
to keep an import used. LightTables::from_parts is used by its tests, so
it is #[cfg(test)] instead of allowed.
Protocol id tables and entity data serializer ids stay: they mirror
vanilla's registries in full. The launcher's three unused items are left
for a launcher PR.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The server sends a knocked-back, pushed or jumping mob's move at once
(Entity.needsSync), and the remote client walks every living entity along
the server's positions with vanilla's SteppedInterpolationHandler: each
position takes the type's update interval to reach, sped up while steps
pile up. Before, the client waited for the next due update and then
lerped there in fewer ticks, so a hit pig jumped away fast.
tools/test_stepped_interpolation_parity.py drives vanilla's unmodified
InterpolationData and the Rust port with the same packets and ticks
(random cases and a punched pig's knockback); 135236 positions and
rotations match bit for bit.
Fixes#1374
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>