ModelTranslator: document quad retopo upgrade
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
This commit is contained in:
ud18010
co-authored by
Claude Opus 4.8
Claude Opus 4.8
parent
b84dd1c53e
commit
25f84f8557
@@ -0,0 +1,433 @@
|
||||
# `--reducer quad` 可靠四边重拓扑升级 Implementation Plan
|
||||
|
||||
> **For agentic workers:** REQUIRED SUB-SKILL: Use superpowers:subagent-driven-development (recommended) or superpowers:executing-plans to implement this plan task-by-task. Steps use checkbox (`- [ ]`) syntax for tracking.
|
||||
|
||||
**Goal:** 把 `--reducer quad` 升级为可靠四边重拓扑：QR 前预 collapse 到中等密度、补小洞、软水密门、保四边不三角化，让 quad 后端在高密度模型上也能产出干净四边低模。
|
||||
|
||||
**Architecture:** 纯策略（`topology_acceptable` 加容忍参数、`precollapse_target`）进 `qr_bridge.py` 可 TDD；Blender 操作（预 collapse、`fill_holes`、保四边导出）进 `bl_decimate.py`，扩展 `_remesh_quad` 与 `main`；`model_decimate.py` 加 `--qr-input-cap` 透传。collapse 后端与默认值零变化。
|
||||
|
||||
**Tech Stack:** Python 3 标准库、Blender 5.0 `bpy`、`unittest`、现有 headless Blender runner。
|
||||
|
||||
设计依据：`docs/superpowers/specs/2026-07-23-quad-reducer-retopo-upgrade-design.md`
|
||||
|
||||
## Global Constraints
|
||||
|
||||
- 不加外部依赖。
|
||||
- `collapse` 后端行为零变化；`--reducer` 默认仍为 `quad`；QR 失败仍自动回退 collapse。
|
||||
- 只 stage 本计划涉及的源码/文档；不碰用户脏工作区（Unity/src/`out_*` 产物）。
|
||||
- 纯逻辑 TDD；Blender 侧 `_remesh_quad`/`main` 改动由真机 smoke 覆盖，本地以 `py_compile`+回归把关。
|
||||
|
||||
## File Map
|
||||
|
||||
- Modify `Tools/ModelTranslator/qr_bridge.py`：`topology_acceptable` 加 `tol`；新增 `precollapse_target` 与 3 个常量。
|
||||
- Modify `Tools/ModelTranslator/tests/test_qr_bridge.py`：软门 tol 与 `precollapse_target` 单测。
|
||||
- Modify `Tools/ModelTranslator/bl_decimate.py`：新增 `_fill_small_holes`；`_remesh_quad` 补洞+软门；`main` 预 collapse+保四边。
|
||||
- Modify `Tools/ModelTranslator/model_decimate.py`：加 `--qr-input-cap` 透传。
|
||||
- Modify `Tools/ModelTranslator/tests/test_model_decimate_cli.py`：适配新增 argv 位。
|
||||
- Modify `Tools/ModelTranslator/README.md`：记录升级后的 quad 行为与 `--qr-input-cap`。
|
||||
|
||||
---
|
||||
|
||||
### Task 1: 纯策略（容忍软门 + 预 collapse 目标）
|
||||
|
||||
**Files:**
|
||||
- Modify: `Tools/ModelTranslator/qr_bridge.py`
|
||||
- Modify: `Tools/ModelTranslator/tests/test_qr_bridge.py`
|
||||
|
||||
- [ ] **Step 1: 写失败单测**
|
||||
|
||||
在 `tests/test_qr_bridge.py` 的 `if __name__ == "__main__":` 之前插入两个测试类：
|
||||
|
||||
```python
|
||||
class TestTopologyTolerance(unittest.TestCase):
|
||||
def test_tol_allows_small_increase(self):
|
||||
# 封闭输入 + QR 多出 3 开边，容忍 12 内 -> 接受
|
||||
self.assertTrue(qb.topology_acceptable(0, 0, 3, 0, tol=12))
|
||||
|
||||
def test_tol_rejects_large_increase(self):
|
||||
self.assertFalse(qb.topology_acceptable(0, 0, 13, 0, tol=12))
|
||||
|
||||
def test_tol_applies_to_nonmanifold(self):
|
||||
self.assertTrue(qb.topology_acceptable(0, 0, 0, 12, tol=12))
|
||||
self.assertFalse(qb.topology_acceptable(0, 0, 0, 13, tol=12))
|
||||
|
||||
def test_default_tol_zero_strict(self):
|
||||
# 默认 tol=0：现有严格语义不变
|
||||
self.assertFalse(qb.topology_acceptable(0, 0, 1, 0))
|
||||
self.assertTrue(qb.topology_acceptable(0, 0, 0, 0))
|
||||
|
||||
|
||||
class TestPrecollapseTarget(unittest.TestCase):
|
||||
def test_above_cap_returns_cap(self):
|
||||
self.assertEqual(qb.precollapse_target(1000000, 50000), 50000)
|
||||
|
||||
def test_below_cap_returns_none(self):
|
||||
self.assertIsNone(qb.precollapse_target(40000, 50000))
|
||||
|
||||
def test_at_cap_returns_none(self):
|
||||
self.assertIsNone(qb.precollapse_target(50000, 50000))
|
||||
|
||||
def test_default_cap(self):
|
||||
self.assertEqual(qb.precollapse_target(60000), qb.QR_INPUT_CAP)
|
||||
self.assertIsNone(qb.precollapse_target(40000))
|
||||
```
|
||||
|
||||
- [ ] **Step 2: 跑测试确认 RED**
|
||||
|
||||
Run（在 `Tools/ModelTranslator`）：
|
||||
```bash
|
||||
python -m unittest tests.test_qr_bridge.TestTopologyTolerance tests.test_qr_bridge.TestPrecollapseTarget -v
|
||||
```
|
||||
Expected: `AttributeError`（`precollapse_target` 不存在）/ `TypeError`（`topology_acceptable` 不接受 `tol`）。
|
||||
|
||||
- [ ] **Step 3: 实现常量 + 函数**
|
||||
|
||||
在 `qr_bridge.py` 顶部常量 `QR_TIMEOUT = 120.0` 之后新增三个常量：
|
||||
```python
|
||||
QR_INPUT_CAP = 50000 # quad 后端 QR 前预 collapse 的中等密度目标面数
|
||||
QR_HOLE_MAX_SIDES = 6 # 补洞上限边数：只补 <=6 边的薄部位小洞
|
||||
QR_CATASTROPHIC_TOL = 12 # 软水密门容忍：补洞后仍多出超此数的开边/非流形才判灾难性
|
||||
```
|
||||
|
||||
把 `topology_acceptable` 整个函数替换为带 `tol` 版本：
|
||||
```python
|
||||
def topology_acceptable(src_boundary, src_nonmanifold,
|
||||
out_boundary, out_nonmanifold, tol=0):
|
||||
"""QR 输出拓扑是否可接受：不比输入多出超过 tol 的开边(孔洞)/非流形边。
|
||||
tol=0（默认）为零容忍严格门；quad 后端补小洞后用 tol=QR_CATASTROPHIC_TOL 只挡灾难性破损。"""
|
||||
return (out_boundary <= src_boundary + tol
|
||||
and out_nonmanifold <= src_nonmanifold + tol)
|
||||
```
|
||||
|
||||
在文件末尾新增：
|
||||
```python
|
||||
def precollapse_target(face_count, cap=QR_INPUT_CAP):
|
||||
"""QR 前预 collapse 目标：面数 > cap 返回 cap（先减到中等密度），否则 None（跳过）。
|
||||
高密度网格直接 QR 必破洞，先 collapse 到中等密度给 QR 一个稳定输入。"""
|
||||
return cap if face_count > cap else None
|
||||
```
|
||||
|
||||
- [ ] **Step 4: 跑测试确认 GREEN + 全量回归**
|
||||
|
||||
Run：
|
||||
```bash
|
||||
python -m unittest tests.test_qr_bridge -v
|
||||
python -m unittest discover -s tests
|
||||
```
|
||||
Expected: 新测试全过；现有 `TestTopologyAcceptable`（4 参调用、tol 默认 0）仍全过；全量零失败。
|
||||
|
||||
- [ ] **Step 5: 提交**
|
||||
|
||||
```bash
|
||||
git add -- Tools/ModelTranslator/qr_bridge.py Tools/ModelTranslator/tests/test_qr_bridge.py
|
||||
git commit -m "ModelTranslator: QR tolerance soft-gate + precollapse target policy"
|
||||
```
|
||||
提交信息末尾附一行：`Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>`
|
||||
|
||||
---
|
||||
|
||||
### Task 2: bl_decimate 预 collapse + 补洞 + 软门 + 保四边
|
||||
|
||||
**Files:**
|
||||
- Modify: `Tools/ModelTranslator/bl_decimate.py`
|
||||
|
||||
> `_remesh_quad`/`main` 依赖 bpy，无法纯单测；本任务以 `py_compile` + 回归把关，真机 smoke 在 Task 4。
|
||||
|
||||
- [ ] **Step 1: 新增 `_fill_small_holes`**
|
||||
|
||||
在 `bl_decimate.py` 的 `_edge_defects` 函数之后插入：
|
||||
```python
|
||||
def _fill_small_holes(obj, max_sides, warnings):
|
||||
"""补掉 <=max_sides 边的小洞（QR 薄部位零星孔洞）；异常降级不中断。"""
|
||||
import bpy
|
||||
try:
|
||||
bpy.ops.object.select_all(action='DESELECT')
|
||||
obj.select_set(True)
|
||||
bpy.context.view_layer.objects.active = obj
|
||||
bpy.ops.object.mode_set(mode='EDIT')
|
||||
bpy.ops.mesh.select_all(action='SELECT')
|
||||
bpy.ops.mesh.fill_holes(sides=max_sides)
|
||||
bpy.ops.object.mode_set(mode='OBJECT')
|
||||
except RuntimeError as e:
|
||||
warnings.append("fill_holes 失败（%s），跳过补洞" % e)
|
||||
```
|
||||
|
||||
- [ ] **Step 2: `_remesh_quad` 导回后补洞 + 软门**
|
||||
|
||||
在 `_remesh_quad` 中，找到这段（导回后的水密门块）：
|
||||
```python
|
||||
retopo = new[0]
|
||||
# 水密门：QR 引擎非确定性地会打出孔洞/非流形，破损即弃用 QR 回退 collapse
|
||||
qr_boundary, qr_nonman = _edge_defects(retopo)
|
||||
if not qr_bridge.topology_acceptable(src_boundary, src_nonman,
|
||||
qr_boundary, qr_nonman):
|
||||
warnings.append("QR 输出拓扑破损（开边 %d→%d，非流形 %d→%d），回退 collapse"
|
||||
% (src_boundary, qr_boundary, src_nonman, qr_nonman))
|
||||
```
|
||||
整体替换为：
|
||||
```python
|
||||
retopo = new[0]
|
||||
# 补小洞：QR 常在薄部位留零星小洞，先补掉再判软门
|
||||
_fill_small_holes(retopo, qr_bridge.QR_HOLE_MAX_SIDES, warnings)
|
||||
# 软水密门：补洞后仍比输入多出超容忍的破损才判灾难性，弃用 QR 回退 collapse
|
||||
qr_boundary, qr_nonman = _edge_defects(retopo)
|
||||
if not qr_bridge.topology_acceptable(src_boundary, src_nonman,
|
||||
qr_boundary, qr_nonman,
|
||||
tol=qr_bridge.QR_CATASTROPHIC_TOL):
|
||||
warnings.append("QR 输出灾难性破损（开边 %d→%d，非流形 %d→%d），回退 collapse"
|
||||
% (src_boundary, qr_boundary, src_nonman, qr_nonman))
|
||||
```
|
||||
（其后的 `bpy.data.objects.remove(retopo...)` / 恢复选中 / return False 与接受分支保持不变。）
|
||||
|
||||
- [ ] **Step 3: `main()` 解析 qr_input_cap**
|
||||
|
||||
在 `main()` 中，找到：
|
||||
```python
|
||||
reducer_arg = argv[5] if len(argv) > 5 else "collapse"
|
||||
warnings = []
|
||||
```
|
||||
改为（其间新增一行）：
|
||||
```python
|
||||
reducer_arg = argv[5] if len(argv) > 5 else "collapse"
|
||||
qr_input_cap = int(argv[6]) if len(argv) > 6 else qr_bridge.QR_INPUT_CAP
|
||||
warnings = []
|
||||
```
|
||||
|
||||
- [ ] **Step 4: `main()` quad 分支预 collapse + 保四边**
|
||||
|
||||
找到 quad 分支：
|
||||
```python
|
||||
if reducer_arg == "quad":
|
||||
if _remesh_quad(obj, target, warnings):
|
||||
obj = bpy.context.view_layer.objects.active # QR 导回的 quad 对象
|
||||
_triangulate(obj) # 三角化供导出/计数
|
||||
cur = len(obj.data.polygons)
|
||||
used_reducer = "quad"
|
||||
did_quad = True
|
||||
else:
|
||||
used_reducer = "quad->collapse"
|
||||
```
|
||||
整体替换为：
|
||||
```python
|
||||
if reducer_arg == "quad":
|
||||
# 预 collapse 到中等密度：给 QR 一个不炸洞的输入（高密度直接 QR 必破）
|
||||
pre = qr_bridge.precollapse_target(len(obj.data.polygons), qr_input_cap)
|
||||
if pre is not None:
|
||||
_decimate(obj, decimate_ratio(pre, len(obj.data.polygons)))
|
||||
_triangulate(obj)
|
||||
if _remesh_quad(obj, target, warnings):
|
||||
obj = bpy.context.view_layer.objects.active # QR 导回的 quad 对象
|
||||
# 保四边：不三角化，直接导出四边网格（FBX/烘焙/Unity 均支持四边）
|
||||
cur = len(obj.data.polygons)
|
||||
used_reducer = "quad"
|
||||
did_quad = True
|
||||
else:
|
||||
used_reducer = "quad->collapse"
|
||||
```
|
||||
（`orig` 仍在此分支之前捕获 = 预 collapse 前的高模面数，作 `tris_before`。）
|
||||
|
||||
- [ ] **Step 5: 更新文件头 argv 文档**
|
||||
|
||||
找到 `bl_decimate.py` 顶部 docstring 里的调用行：
|
||||
```
|
||||
<src.fbx> <out.fbx> <target_tris> [unwrap:seam|smart] [overlap_tol] [reducer:collapse|quad]
|
||||
```
|
||||
改为：
|
||||
```
|
||||
<src.fbx> <out.fbx> <target_tris> [unwrap:seam|smart] [overlap_tol] [reducer:collapse|quad] [qr_input_cap]
|
||||
```
|
||||
|
||||
- [ ] **Step 6: 编译 + 回归**
|
||||
|
||||
Run（在 `Tools/ModelTranslator`）：
|
||||
```bash
|
||||
python -m py_compile qr_bridge.py bl_decimate.py
|
||||
python -m unittest discover -s tests
|
||||
```
|
||||
Expected: 编译 exit 0；全部测试通过（本任务未改纯函数与 CLI 测试，回归应全绿）。
|
||||
|
||||
- [ ] **Step 7: 提交**
|
||||
|
||||
```bash
|
||||
git add -- Tools/ModelTranslator/bl_decimate.py
|
||||
git commit -m "ModelTranslator: quad backend precollapse + fill-holes + keep-quads"
|
||||
```
|
||||
提交信息末尾附一行：`Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>`
|
||||
|
||||
---
|
||||
|
||||
### Task 3: model_decimate 加 `--qr-input-cap`（并修 CLI 测试）
|
||||
|
||||
**Files:**
|
||||
- Modify: `Tools/ModelTranslator/model_decimate.py`
|
||||
- Modify: `Tools/ModelTranslator/tests/test_model_decimate_cli.py`
|
||||
|
||||
- [ ] **Step 1: 先改 CLI 测试为 RED（反映新增 argv 位）**
|
||||
|
||||
现有 `tests/test_model_decimate_cli.py` 断言 `args[-1] == "quad"`；新增 `--qr-input-cap` 后 reducer 不再是末位。把该文件里的断言块：
|
||||
```python
|
||||
args = run.call_args.args[2]
|
||||
self.assertEqual(args[-1], "quad")
|
||||
```
|
||||
替换为：
|
||||
```python
|
||||
args = run.call_args.args[2]
|
||||
# argv 顺序：input, out, tris, unwrap, overlap, reducer, qr_input_cap
|
||||
self.assertEqual(args[5], "quad") # reducer 默认 quad
|
||||
self.assertEqual(args[6], "50000") # qr_input_cap 默认 50000
|
||||
```
|
||||
|
||||
- [ ] **Step 2: 跑测试确认 RED**
|
||||
|
||||
Run：
|
||||
```bash
|
||||
python -m unittest tests.test_model_decimate_cli -v
|
||||
```
|
||||
Expected: FAIL（当前 `model_decimate` 只传 6 个 argv，`args[6]` 越界 `IndexError`）。
|
||||
|
||||
- [ ] **Step 3: 加 `--qr-input-cap` 并透传**
|
||||
|
||||
在 `model_decimate.py` 顶部 import 处：
|
||||
```python
|
||||
from mt_run import find_blender, run_blender_script
|
||||
```
|
||||
其后新增：
|
||||
```python
|
||||
from qr_bridge import QR_INPUT_CAP
|
||||
```
|
||||
|
||||
在 `--reducer` 参数定义之后新增：
|
||||
```python
|
||||
ap.add_argument("--qr-input-cap", type=int, default=QR_INPUT_CAP,
|
||||
help="quad 后端 QR 前预 collapse 的中等密度目标面数（默认 50000）；高密度模型直接 QR 会破洞")
|
||||
```
|
||||
|
||||
在 `args = ap.parse_args()` 后的校验区新增：
|
||||
```python
|
||||
if args.qr_input_cap <= 0:
|
||||
ap.error("--qr-input-cap 必须为正整数")
|
||||
```
|
||||
|
||||
把 `run_blender_script` 调用：
|
||||
```python
|
||||
s = run_blender_script(blender, "bl_decimate.py",
|
||||
[args.input, out_fbx, str(args.tris), args.unwrap,
|
||||
"" if args.max_overlap is None else str(args.max_overlap),
|
||||
args.reducer])
|
||||
```
|
||||
改为（追加 `str(args.qr_input_cap)`）：
|
||||
```python
|
||||
s = run_blender_script(blender, "bl_decimate.py",
|
||||
[args.input, out_fbx, str(args.tris), args.unwrap,
|
||||
"" if args.max_overlap is None else str(args.max_overlap),
|
||||
args.reducer, str(args.qr_input_cap)])
|
||||
```
|
||||
|
||||
- [ ] **Step 4: 跑测试确认 GREEN + 编译 + --help**
|
||||
|
||||
Run：
|
||||
```bash
|
||||
python -m py_compile model_decimate.py
|
||||
python -m unittest tests.test_model_decimate_cli -v
|
||||
python model_decimate.py --help
|
||||
```
|
||||
Expected: 编译 exit 0；CLI 测试通过；`--help` 含 `--qr-input-cap`。
|
||||
|
||||
- [ ] **Step 5: 提交**
|
||||
|
||||
```bash
|
||||
git add -- Tools/ModelTranslator/model_decimate.py Tools/ModelTranslator/tests/test_model_decimate_cli.py
|
||||
git commit -m "ModelTranslator: expose --qr-input-cap, adapt CLI test to new argv"
|
||||
```
|
||||
提交信息末尾附一行：`Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>`
|
||||
|
||||
---
|
||||
|
||||
### Task 4: 真机验证与文档
|
||||
|
||||
**Files:**
|
||||
- Modify: `Tools/ModelTranslator/README.md`
|
||||
|
||||
Blender 路径：`D:/tools/Blender/blender-5.0.0-windows-x64/blender.exe`（用 `--blender` 显式传）。
|
||||
|
||||
- [ ] **Step 1: quad 后端真机 smoke（有机角色，验证保四边）**
|
||||
|
||||
Run（`out_*` 已被 gitignore，勿提交产物）：
|
||||
```bash
|
||||
python model_decimate.py "C:/Users/Administrator/Downloads/supergirl.fbx" --tris 10000 --reducer quad -o out_sg_test --blender "D:/tools/Blender/blender-5.0.0-windows-x64/blender.exe"
|
||||
```
|
||||
Expected：打印 `后端: quad`；`out_sg_test/supergirl_low.fbx` 存在。用下面命令验证输出以四边为主、缺陷少：
|
||||
```bash
|
||||
"D:/tools/Blender/blender-5.0.0-windows-x64/blender.exe" -b --factory-startup --python-expr "
|
||||
import bpy,bmesh
|
||||
bpy.ops.wm.read_factory_settings(use_empty=True)
|
||||
bpy.ops.import_scene.fbx(filepath=r'D:/UD/AI/AIC#Project/Tools/ModelTranslator/out_sg_test/supergirl_low.fbx')
|
||||
o=[x for x in bpy.data.objects if x.type=='MESH'][0]
|
||||
bm=bmesh.new();bm.from_mesh(o.data)
|
||||
q=sum(1 for f in bm.faces if len(f.verts)==4);t=sum(1 for f in bm.faces if len(f.verts)==3)
|
||||
b=sum(1 for e in bm.edges if len(e.link_faces)==1);nm=sum(1 for e in bm.edges if len(e.link_faces)>2)
|
||||
print('CHECK faces=%d quad=%d tri=%d boundary=%d nonmanifold=%d'%(len(bm.faces),q,t,b,nm))
|
||||
"
|
||||
```
|
||||
Expected：`quad` 数 >> `tri` 数（以四边为主）；`boundary`/`nonmanifold` 为个位数小值（补洞 + 软门生效）。
|
||||
|
||||
- [ ] **Step 2: collapse 后端回归（零变化）**
|
||||
|
||||
Run：
|
||||
```bash
|
||||
python model_decimate.py "C:/Users/Administrator/Downloads/supergirl.fbx" --tris 10000 --reducer collapse -o out_sg_collapse --blender "D:/tools/Blender/blender-5.0.0-windows-x64/blender.exe"
|
||||
```
|
||||
Expected：打印 `后端: collapse`；面数正好 10000（±3%）；输出全三角（collapse 行为与升级前一致）。
|
||||
|
||||
- [ ] **Step 3: 强制回退验证**
|
||||
|
||||
Run：
|
||||
```bash
|
||||
QR_ENGINE="/nonexistent/xremesh.exe" python model_decimate.py "C:/Users/Administrator/Downloads/supergirl.fbx" --tris 10000 --reducer quad -o out_sg_fb --blender "D:/tools/Blender/blender-5.0.0-windows-x64/blender.exe"
|
||||
```
|
||||
Expected：打印 `后端: quad->collapse`；输出正常生成；warnings 含"未找到 QuadRemesher 引擎"。
|
||||
|
||||
- [ ] **Step 4: 更新 README**
|
||||
|
||||
在 `Tools/ModelTranslator/README.md` 的 `--reducer` 段落，把 quad 的回退描述扩充为升级后行为。找到该 bullet 中的这句：
|
||||
```markdown
|
||||
**水密门**：QR 引擎在激进减面下会*非确定性*地打出孔洞/非流形边（同一封闭输入实测约 85% 次数破碎），故导回后比对输入拓扑——QR 若比输入新增开边(孔洞)/非流形边即判破损、弃用 QR 回退 collapse（保证减面输出的拓扑完整性不劣于源模，杜绝"破碎面"）。
|
||||
```
|
||||
替换为：
|
||||
```markdown
|
||||
**升级流程**：quad 后端先把高密度输入**预 collapse 到中等密度**（`--qr-input-cap`，默认 50000——高密度网格直接 QR 必破洞），再 QR，导回后 `fill_holes` 补掉薄部位小洞，最后**软水密门**（补洞后仍比输入多出超容忍的破损才判灾难性、回退 collapse）。quad 成功时**保留四边、不三角化**导出（FBX/烘焙/Unity 均支持四边；这修掉了旧版 quad 被三角化的问题）。
|
||||
```
|
||||
|
||||
- [ ] **Step 5: 最终自动化验证**
|
||||
|
||||
Run：
|
||||
```bash
|
||||
python -m py_compile qr_bridge.py bl_decimate.py model_decimate.py
|
||||
python -m unittest discover -s tests
|
||||
```
|
||||
Expected：编译 exit 0；全部测试通过、零失败。
|
||||
|
||||
- [ ] **Step 6: 检查 diff 与暂存范围**
|
||||
|
||||
Run：
|
||||
```bash
|
||||
git status --short
|
||||
git diff --check
|
||||
```
|
||||
Expected：无空白错误；`out_*` 产物未被 stage；用户既有脏文件（unity_assets.py 等）保持未暂存。
|
||||
|
||||
- [ ] **Step 7: 提交文档**
|
||||
|
||||
```bash
|
||||
git add -- Tools/ModelTranslator/README.md docs/superpowers/specs/2026-07-23-quad-reducer-retopo-upgrade-design.md docs/superpowers/plans/2026-07-23-quad-reducer-retopo-upgrade.md
|
||||
git commit -m "ModelTranslator: document quad retopo upgrade"
|
||||
```
|
||||
提交信息末尾附一行：`Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>`
|
||||
|
||||
---
|
||||
|
||||
## Self-Review 记录
|
||||
|
||||
- **Spec 覆盖**：预 collapse（Task2 Step4 + Task1 `precollapse_target`）、补小洞（Task2 Step1/2）、软门（Task1 tol + Task2 Step2）、保四边不三角化（Task2 Step4 去掉 `_triangulate`）、`--qr-input-cap`（Task3）、水密门保底/回退（Task2 Step2 软门 + 现有回退）、collapse 零变化（未动 collapse 分支）、测试（Task1 单测 + Task3 CLI 测试 + Task4 真机）、README（Task4）——全覆盖。
|
||||
- **占位符**：无 TBD/TODO；所有代码步骤含完整代码与锚点。
|
||||
- **类型一致**：`topology_acceptable(...tol=0)` 定义（Task1）与 `_remesh_quad` 调用 `tol=qr_bridge.QR_CATASTROPHIC_TOL`（Task2）一致；`precollapse_target` 定义（Task1）与 main 调用（Task2）签名一致；argv 第 7 位 `qr_input_cap` 在 bl_decimate（Task2 Step3 `argv[6]`）、model_decimate（Task3 追加第 7 位）、CLI 测试（Task3 `args[6]`）三处口径一致；常量 `QR_INPUT_CAP`/`QR_HOLE_MAX_SIDES`/`QR_CATASTROPHIC_TOL` 定义（Task1）与消费（Task2/Task3）一致。
|
||||
Reference in New Issue
Block a user