# WASM Regression Tests

## How to run

1. Build the business WASM module:

```powershell
npm run build
```

2. Run the regression suite:

```powershell
npm run test:wasm
```

3. Build and run in one step:

```powershell
npm test
```

The runner loads `public/wasm/smart_math.js`, calls `init_func`, then executes the business API through `func`.

## Test program

- Runner: `scripts/run_wasm_tests.js`
- Entry exports used:
  - `_init_func`
  - `_func`
  - `_smart_free_string`

The runner is data-driven for stable request samples, and adds two dynamic FK/IK roundtrip cases so inverse kinematics is checked against forward kinematics instead of hard-coded pose strings.

## Test data

### Kinematics

- `tests/testdata/kinematics/list_robots.json`
  - Verifies the built-in robot initialization completed and at least two robots are available.
- `tests/testdata/kinematics/forward_zero.json`
  - Verifies TCP forward kinematics at zero joints.
- `tests/testdata/kinematics/forward_all_joints_path.json`
  - Verifies all-joint forward export for two frames.
- `tests/testdata/kinematics/roundtrip_single_seed.json`
  - Seed joints for FK -> IK -> FK single-pose regression.
- `tests/testdata/kinematics/roundtrip_path_seed.json`
  - Seed joints for two-pose path interpolation regression.

### SPC

- `tests/testdata/spc/basic_5x30.json`
  - Canonical 150-point SPC sample.
  - Uses the actual command name `Cmd_Spc`.

### Four-bar / Crank slider

- `tests/testdata/fourbar/crank_slider_valid.json`
  - Normal calculation case.
- `tests/testdata/fourbar/crank_slider_invalid.json`
  - Invalid parameter case to verify validation and error return.

### Quadruped robot

- `tests/testdata/quadruped/calculate_points_from_motor_angles.json`
  - Verifies reverse result export from motor angles.
- `tests/testdata/quadruped/perform_forward_kinematics.json`
  - Verifies gait trajectory generation with explicit gait/system parameters.

## What is asserted

- Top-level API success and required result fields.
- Expected frame count / robot count / array length.
- Four-bar validation errors for bad inputs.
- SPC subgroup size and spec limits.
- Quadruped result structure.
- FK/IK roundtrip tolerances:
  - single pose position tolerance: `0.002`
  - single pose quaternion tolerance: `0.002`
  - path endpoint position tolerance: `0.01`

## Notes

- The old ad-hoc files such as `tests/fixtures/legacy/spc_in.txt` and `tests/fixtures/legacy/WorkTest.json` are still useful as raw references, but the regression runner uses the normalized data under `tests/testdata`.
- If you change URDF, robot naming, or quadruped default config, update the corresponding files under `tests/testdata` first.