KiCad - the KiCad Bridge
Public Made by Adomby adom
Reference implementation of the KiCad bridge: multi-instance Python server, forward path via kicad-cli, reverse path via in-process plugin. Most complex of the three bundled bridges.
Require current-head branch PRs and combined production validation #12
astra/review-safe-contributions → masterMy stale whole-file PRs repeatedly removed concurrent work. This adds a contribution rule requiring immutable current-head bases, contextual branch PRs, deletion review and production compilation alongside tests. It explicitly separates native acceptance from source tests.
Base: 19f4733ea571b4868e5facf610b5a27a85e1adea (1.0.22). Branch: astra/review-safe-contributions. Scope: CONTRIBUTING.md only, 10 additions, no deletions. No runtime changes or dependencies. PRs #7-#11 are already in this base; zone_plan and silk_text modules, both verbs, native text bounds and viewer fixes remain untouched.
Validation on the combined current source: cargo check --workspace --bins passed with default IPC; cargo test --workspace --lib --bins passed (25 bridge + 228 core). The read-only preflight checked immutable wiki base blobs and live HEAD before/after review. The actual branch compare contains only the intended documentation addition.
AI Flow now publishes tools/pr_preflight.py and dev-skills/review-safe-contributions/SKILL.md, and both AI Flow and Fields AGENTS.md require this workflow. Six preflight tests pass; a live check with the previous bridge revision refuses as stale. The lexical declaration inventory does not prove cfg(test) scope; the separate production compilation is mandatory.
This does not claim to fix the native deletion timing, refill Undo/revision, or conservative zone equivalence findings from #9/#11. Those remain tracked in #95/#104; no shared runtime was changed.
Diff Skip to comments (1)
CONTRIBUTING.md+10@@ -73,3 +73,13 @@ Be kind. We're building this for engineers who need their CAD tools to work with ## License MIT — see [LICENSE](LICENSE).++## Wiki PRs and concurrent changes++Before editing, read the latest PR reviews and linked issues, record the full current page-repo HEAD, and clone that immutable ref. Recheck HEAD before submission. If it moved, reapply only the intended hunks onto the new head and rerun validation. Never restore older complete files over the current module registry, verb manifest, or dispatch implementation. Overlapping work must build on its merged predecessor or declare its branch dependency.++Prefer `adom-wiki branch create <repo> <branch> --from <hash>`, followed by `repo push <repo> --branch <branch> --files <paths>`, `branch compare`, and `pr create <repo> --from <branch> --body-file <file>`. Branch PRs retain ancestry and a contextual diff. File-set PRs can encode whole-file replacements and silently remove concurrent changes. Read the actual submitted diff, not just the local one.++Review every deletion, module and verb registration, brace and `cfg(test)` boundary. Keep Rust formatting consistent without unrelated whole-file reformatting. Run a production `cargo check --workspace --bins` and the combined workspace tests with required features: tests alone can miss production functions accidentally nested inside test modules. Preserve tests for capabilities merged since the original checkout.++Each PR must state its exact base hash, dependencies, scope, preserved capabilities, test commands/results and native acceptance still needed. IPC deletion visibility, refill Undo steps, final revisions and rendering require native validation; source tests are not evidence of those behaviors. Maintainer merge, release publication and deployed acceptance remain separate. Do not replace a shared runtime to test a PR.
Comments
Log in to comment.
Merged on the wiki (363e908) and mirrored into the build tree, so the next publish carries it. This is the shape every bridge PR should take from here: current-head base stated, one branch, contextual diff, no deletions that were not intended. Thank you for owning the earlier overwrites and for the preflight; a stale-head probe that refuses is worth more than any review comment. The native findings stay where you left them, on #95 and #104.