Refactor CI pipeline scripts and documentation

- Consolidated CI pipeline steps into PowerShell and Bash scripts for improved usability and consistency.
- Updated documentation in `ci.md` to clarify the execution process and options for skipping the rbx-dom section.
- Enhanced error handling and reporting in the scripts to provide clearer feedback on CI step results.
- Streamlined formatting and linting processes for both Rojo and rbx-dom, ensuring a more efficient CI workflow.
This commit is contained in:
Joe 2026-02-10 16:46:26 -06:00
parent 9688224cb4
commit f496bee347
8 changed files with 473 additions and 193 deletions

View file

@ -1,185 +1,50 @@
# /ci - Complete CI Pipeline
Run the complete CI pipeline: auto-fix all formatting, run all linters, build everything, and run ALL tests.
Run the complete CI pipeline by executing the appropriate platform script.
## Instructions
- If the user's message contains **"rbx-dom"**, run **both** the Rojo and rbx-dom sections below.
- Otherwise, **skip the entire rbx-dom section** (steps 9–14) to save time.
### 1. Determine rbx-dom Flag
Execute ALL applicable steps in order. Fix issues automatically where possible.
- By default, run **without** the skip flag (full suite including rbx-dom).
- If the user's message asks to **skip rbx-dom**, pass the skip flag (`-NoRbxDom` / `--no-rbx-dom`).
---
### 2. Run the CI Script
## Rojo
### 1. Auto-fix Lua Formatting (Stylua)
**Windows (PowerShell):**
```powershell
stylua plugin/src
.\scripts\ci.ps1 # with rbx-dom
.\scripts\ci.ps1 -NoRbxDom # without rbx-dom
```
### 2. Auto-fix Rust Formatting
```powershell
cargo fmt
**macOS/Linux (Bash):**
```bash
bash scripts/ci.sh # with rbx-dom
bash scripts/ci.sh --no-rbx-dom # without rbx-dom
```
### 3. Auto-fix TypeScript Formatting (Prettier) for vscode-rojo
```powershell
cd vscode-rojo
npx prettier --write "src/**/*.ts"
cd ..
```
**Monitor the script output as it runs.** After each step completes, check its result. If a step failed, kill the script and switch to manual fixing (step 3) — don't let it continue to the next step.
### 4. Lua Static Analysis (Selene) - MUST PASS WITH ZERO ISSUES
```powershell
selene plugin/src
```
### 3. Fix Failures Between Steps
**CRITICAL: Selene must exit with code 0 (zero errors AND zero warnings).** Warnings count as failures.
When a step fails, stop the script and fix the issue before continuing. Use targeted commands to speed up the feedback loop:
If Selene reports ANY errors or warnings:
1. **FIX THEM IMMEDIATELY** before proceeding - do not just report them
2. Common fixes:
- `unused_variable`: Replace with `_` (e.g., `for _ = 1, 10 do`) or remove the variable entirely
- `unused_variable` for imports: Remove the unused require/import line
3. Re-run Selene after fixes to confirm exit code 0
4. Only proceed once Selene shows: `Results: 0 errors, 0 warnings, 0 parse errors`
1. **Formatting/linting failures** — auto-fix with the relevant formatter (`cargo fmt`, `stylua`), then verify with the specific check command only (e.g. `cargo fmt --check`).
### 5. Rust Linting (Clippy) - Auto-fix where possible
```powershell
cargo clippy --fix --allow-dirty --allow-staged
```
Then verify with:
```powershell
cargo clippy --all-targets --all-features
```
If there are remaining warnings/errors, fix them before proceeding.
2. **Test failures — investigate the implementation, not just the tests.** For every failing test:
- Read the failing test to understand what behavior it asserts.
- Read the implementation code the test exercises.
- Determine the root cause:
- **Implementation regression/bug:** The feature is incomplete or broken. **Fix the implementation** so the test passes. Do not weaken or delete the test.
- **Outdated test fixture:** The test's snapshot or fixture data is stale because of an intentional change. Update the fixture with `cargo insta review`/`cargo insta accept`, or regenerate the fixture data. Confirm the new fixture reflects correct behavior.
- **Incomplete feature:** A new feature was added but doesn't fully handle all cases the tests cover. **Complete the feature implementation** rather than trimming the tests to match a partial implementation.
- **Never** delete or gut a test just to make CI green. Tests exist for a reason — the code must satisfy them, not the other way around.
- **Re-run only the failing test(s)** to verify fixes quickly (e.g. `cargo test <test_name>`, `cargo test -p <crate>`). Do not re-run the entire suite until you believe all issues are resolved.
### 6. Build Everything
```powershell
cargo build --locked --all-targets --all-features
```
3. **After all known issues are fixed**, re-run the full CI script from step 2 to catch any remaining failures.
### 7. Run ALL Rust Tests
```powershell
cargo test --locked --all-targets --all-features
```
4. Repeat until the script exits with `Overall: PASS`.
### 8. Run Roblox Plugin Tests
### 4. Report
**Do NOT use the scripts (`unit-test-plugin.ps1` / `unit-test-plugin.sh`).** They reference a system-installed binary which may not exist. Always use the freshly built binary from step 6.
```powershell
.\target\debug\atlas.exe build plugin/test-place.project.json -o TestPlace.rbxl
run-in-roblox --script plugin/run-tests.server.lua --place TestPlace.rbxl
```
---
## rbx-dom (submodule)
All rbx-dom steps run inside the `rbx-dom/` directory.
### 9. Auto-fix Lua Formatting (Stylua) for rbx_dom_lua
```powershell
cd rbx-dom
stylua rbx_dom_lua/src
cd ..
```
### 10. Auto-fix Rust Formatting
```powershell
cd rbx-dom
cargo fmt
cd ..
```
### 11. Lua Static Analysis (Selene) for rbx_dom_lua - MUST PASS WITH ZERO ISSUES
```powershell
cd rbx-dom
selene rbx_dom_lua/src
cd ..
```
Same rules as step 4: exit code 0 required, fix all errors and warnings before proceeding.
### 12. Rust Linting (Clippy) - Auto-fix where possible
```powershell
cd rbx-dom
cargo clippy --fix --allow-dirty --allow-staged
```
Then verify with:
```powershell
cd rbx-dom
cargo clippy --all-targets --all-features -- -D warnings
cd ..
```
If there are remaining warnings/errors, fix them before proceeding.
### 13. Build rbx-dom
```powershell
cd rbx-dom
cargo build --verbose
cargo build --all-features --verbose
cd ..
```
### 14. Run ALL rbx-dom Rust Tests
```powershell
cd rbx-dom
cargo test --verbose
cargo test --all-features --verbose
cd ..
```
---
## Reporting
After ALL steps complete, provide a summary:
```
=== CI COMPLETE ===
--- Rojo ---
Formatting:
- Stylua: [fixed X files / no changes needed]
- Rustfmt: [fixed X files / no changes needed]
- Prettier (vscode-rojo): [fixed X files / no changes needed]
Linting:
- Selene: PASS (0 errors, 0 warnings) / FAIL (X errors, Y warnings)
- Clippy: PASS / FAIL (X warnings)
Build: PASS / FAIL
Tests:
- Rust: X passed, Y failed
- Plugin: X passed, Y failed
--- rbx-dom (only if included) ---
Formatting:
- Stylua (rbx_dom_lua): [fixed X files / no changes needed]
- Rustfmt: [fixed X files / no changes needed]
Linting:
- Selene (rbx_dom_lua): PASS (0 errors, 0 warnings) / FAIL (X errors, Y warnings)
- Clippy: PASS / FAIL (X warnings)
Build: PASS / FAIL
Build (all features): PASS / FAIL
Tests:
- Rust: X passed, Y failed
- Rust (all features): X passed, Y failed
--- Overall: PASS / FAIL ---
```
**IMPORTANT:**
- Selene PASS requires exit code 0 (0 errors AND 0 warnings). Any warnings = FAIL.
- If any step fails, continue running the remaining steps to get a complete picture, then report all failures at the end.
- Do not report "Overall: PASS" unless ALL linters pass with zero issues across all sections that were run.
- If rbx-dom was skipped, omit its section from the report entirely.
Once CI passes clean, provide a brief summary of what was fixed (if anything) and confirm the final result.