valentineus/fparkan
1
Fork 0
Code Issues Pull Requests 1 Actions Releases Activity

feat: close stage 0-2 audit groundwork

Browse Source 
Remove legacy SDL/OpenGL adapters from the workspace and introduce winit/Vulkan adapter boundaries for the rendered composition root.

Add reproducible toolchain and xtask CI coverage for formatting, tests, clippy, docs, policy, deny, acceptance auditing, and hosted OS matrix evidence.

Strengthen Stage 1 data contracts with byte-first paths, VFS hardening, structured diagnostics, RsLi writer/edit scaffolding, corpus reporting, and resource error classification.

Advance Stage 2 asset preparation by moving mission loading through assets/runtime boundaries, materializing prototype graph data, preserving provenance, and adding inspection/viewer integration.

Record the Stage 0-2 audit input, acceptance roadmap, coverage updates, and documentation notes for follow-up evidence.
This commit is contained in:
Valentin Popov
2026-06-23 22:05:16 +04:00
parent 83d763dd70
commit f8e447ffee
40 changed files with 5323 additions and 1342 deletions
Download Patch File Download Diff File Expand all files Collapse all files
parity/README.md
+11 -8
View File
@@ -1,20 +1,23 @@
# Render Parity Dataset
This folder stores parity-test input for `crates/render-parity`.
This folder stores parity-test input for legacy render comparison workflows.
- `cases.toml`: list of deterministic render cases.
- `reference/*.png`: baseline frames captured from the original renderer.
Expected workflow:
1. Capture baseline PNG frames from original game/editor for each case.
1. Capture baseline PNG frames for each case.
2. Add entries to `cases.toml`.
3. Run:
3. Run the acceptance renderer capture workflow with fixed profiles and compare
output captures out-of-tree.
```bash
cargo run -p render-parity -- \
--manifest parity/cases.toml \
--output-dir target/render-parity/current
```text
1) Prepare `cases.toml` and baseline captures.
2) Run `fparkan-game` (or dedicated acceptance runner) with fixed seed.
3) Compare outputs against baseline in dedicated comparison tooling.
```
On failure, diff images are saved to `target/render-parity/current/diff`.
The `render-parity` crate is no longer present as a standalone runner in this
workspace snapshot; parity evidence is now produced through the acceptance
artifacts and stage audit tooling.