1.8.0-SNAPSHOT: 移动与寻路修复及改进#8
Merged
Merged
Conversation
- Fix velocity lost-update race in updateMotionTask: physics tick now works on a copy and merges its writes atomically instead of overwriting the whole vector, so WalkMovement.onStop / jump impulses / stop are no longer eaten - Fix dimension registry index misalignment in RegistryDataListener (null-data entries now keep their slot) and guard missing min_y/height keys - Follow: stay put within a keep radius instead of repathing every tick; default radius 1 block, configurable via new PathMovement(path, radius) overload and optional `follow <entityId> [keepRadius]` argument - GapJumpStrategy: add diagonal gap jump edges (1-wide diagonal hole) - PathMovement: sprint (1.3x) during gap jump segments so long jumps reach; log gap jump segments once for diagnostics - Goto: tolerate commas in coordinates, allow y up to maxWorldY+1, report partial/no-path results instead of failing silently - Pathfinding without blocks in inventory now skips BridgePillarStrategy via new PathfindingContextBuilder.addNoPlacementStrategies() - Demote per-repath "Path NOT found" warning to debug Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Mid-flight the current target node hangs over the gap, so checkAndPlace would see non-solid ground, find no blocks in the hotbar and kill the whole path in the middle of the jump. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Edges now carry the MovementType that traverses them, and findPath returns List<PathStep> (node + how to reach it) instead of bare nodes. PathMovement executes the type declared by the planner rather than re-deriving it geometrically; dig/bridge/pillar execution moved into BuiltinMovementType.createMovement. Plugins can implement MovementType (with requiresGroundToDispatch for mid-air types) and attach it to edges from their own strategies. A traversability check on built-in edges replans when the world changed since planning. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The per-tick "is the target still walkable?" check replanned on every failure, and a replan resets currentIndex to 0 — whose first target is the centre of the bot's current block. Mid-step that pulls the bot backwards, and isStillTraversable's live isSolid/isPassable probes false-negative often enough (cross-section air reads, non-full ground) to fire constantly, so the bot kept backing up while walking. Replace it with progress-based stuck detection: only replan after ~2s of no horizontal progress toward the target while on the ground. Normal walking always makes progress, so it never triggers and never yanks the bot back, while a genuinely blocked path still recovers. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
ChunkDataListener fires requestRepath() on every nearby block/section update, and repathInternally() rebuilt the path with currentIndex reset to 0. Two failures resulted: - Backing up: newPath[0] is the block the bot already stands in, so after any repath the bot first walked back to that block's centre. With frequent updates it kept getting yanked back a node. - Drifting over gaps: a repath mid jump/fall rebuilt from a mid-air point, dropping the sprint/jump momentum, so the bot floated across a gap instead of clearing it. Defer external repaths until the bot is back on the ground, and start the new path at index 1 (skip the block we're standing in) so the bot heads for the next node — and a gap-jump edge stays the live target. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
findBlock() passed an item id straight into isSolidBlock(int), which expects a block-state id — two unrelated registries. Empty/air slots and random items could land on a solid block's state range, so an empty hotbar was reported as holding placeable blocks. The planner then built bridge/pillar paths it couldn't execute, spamming "NO BLOCKS" and walking the bot into gaps. Resolve the item id to its registry name via ItemRegistry, then look the block up by name with the new BlockStateParser.isSolidBlockByName. Air and non-block items now correctly report no placeable blocks. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
If a bridge/pillar edge's createMovement returns null (e.g. the hotbar ran out of blocks mid-path), the follower used to fall through to plain walking and step straight into the gap the placement was meant to fill, spamming the no-blocks warning each tick. Add MovementType.canWalkWhenNoMovement (true for walk-like moves, false for BRIDGE/PILLAR). When a non-walkable type yields no movement, replan: with an empty hotbar the planner drops placement strategies and routes around the gap or reports the goal unreachable. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
As the bot moves, chunks load and unload and nearby blocks update, firing a burst of requestRepath() calls. Each repath rebuilds the path from the current block, and the planner may pick a different equivalent route, so the bot jittered and stepped backwards a little whenever a new chunk loaded. Stop ForgetLevelChunk (chunk unload) from requesting a repath — it never invalidates the path being walked — and throttle remaining external repath requests to at most one per 20 ticks. Genuine path failures still replan promptly via stuck detection and goal/placement handling. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
DigBlockMovement only computed the dig side; it never aimed at the block, so the bot kept its pre-dig heading (pitch forced to 0). Servers that validate line of sight would reject the dig, and blocks overhead or underfoot were never looked at. Add directLookAt to the block centre in init(), mirroring PlaceBlockMovement. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
On a published release, dispatch an update_version event to the MovementSync_documention repo so its version.json reflects the new release. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
On merging a PR into main, flip the project version from -SNAPSHOT to -RELEASE, build the shaded JARs, and create a draft GitHub Release with them attached, then delete the PR head branch. Mirrors xinbot's flow. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
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
1 participant
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.
修复
updateMotionTask的 velocity 丢失更新竞态——物理 tick 改为操作副本并原子合并写回，不再覆盖并发的WalkMovement.onStop扣减、跳跃冲量和停止指令RegistryDataListener解析 dimension_type registry 时 data 为 null 的条目不再被跳过（known packs 下会导致把下界的 min_y 用到主世界），并对缺失的 min_y/height 键做回退改进
PathMovement(path, radius)重载与follow <entityId> [keepRadius]可选参数可自定义GapJumpStrategy新增斜向跳跃边（跨 1 格宽斜向坑，落点 2 格对角）addNoPlacementStrategies()，背包没有方块时寻路不再考虑搭桥/垫高边（同时省去每次节点扩展扫背包）🤖 Generated with Claude Code