Initial commit -- did OMO do a good job?

2026-09-01 18:29:46 -04:00
commit 20a6c26f7c
179 changed files with 21704 additions and 0 deletions
@@ -0,0 +1,76 @@
+Todo 20 — multi-stage Dockerfile, WASM bundle-size gate, NEWS_DATABASE_URL support
+
+Completed: 2026-09-01
+
+Scope
+- Added `/home/connor/docs/projects/news/Dockerfile` with three stages mirroring
+  `runway/Dockerfile` exactly:
+  - `web`: `rust:1.89-slim-bookworm`, Node 22 copied from a pinned
+    `node:22-bookworm-slim` image, Trunk 0.21.14 installed as a released binary,
+    `npm ci` at the workspace root, then `trunk build --release` in
+    `crates/news-web`.
+  - `server`: `cargo build --release --locked -p news-server`, binary copied
+    out of the cache mount to `/news-server`.
+  - `runtime`: `debian:bookworm-slim` with only `ca-certificates` and `tzdata`,
+    server binary at `/usr/local/bin/news-server`, frontend dist copied to
+    `/srv/dist`, `/db` created, `NEWS_DATABASE_URL=sqlite:///db/news.db`,
+    `EXPOSE 3000`, `CMD ["news-server"]`.
+- Added `.dockerignore` mirroring runway's, with `crates/news-web/dist` and
+  `crates/news-web/styles/generated.css` excluded.
+- Wired `NEWS_DATABASE_URL` as the primary database path source in
+  `crates/news-server/src/bin_support.rs`, stripping the `sqlite://` prefix to
+  obtain the filesystem path while preserving the existing create-if-missing
+  behavior. Falls back to `NEWS_DB_PATH`, then `./news.db`. Existing behavior
+  without the new variable is unchanged.
+- Added a clap `version` flag (`#[command(version)]`) to `news-server` so the
+  container can be smoke-tested with `news-server --version`.
+- Replaced the stub `bundle` job in `.gitea/workflows/ci.yml` with the runway-
+  matching job: install Trunk, `npm ci`, `trunk build --release` in
+  `crates/news-web`, then a 1.8 MiB WASM size assertion that fails with the
+  byte overage printed.
+
+Verification
+- `cargo fmt --all --check` exit 0.
+- `cargo clippy --workspace --all-targets -- -D warnings` exit 0.
+- `cargo test --workspace` green (89 tests: 87 existing + 2 new
+  `bin_support::tests` for `NEWS_DATABASE_URL` parsing).
+
+Container acceptance
+- `podman build -t news-triage:test /home/connor/docs/projects/news` exited 0.
+- `podman run --rm news-triage:test news-server --version` output:
+  `news-server 0.1.0` (no missing shared library error).
+
+QA happy path
+- `podman run --rm news-triage:test ls /srv/dist` output:
+  ```
+  generated-f134b848b9e82d1f.css
+  index.html
+  news-web-8411e96eeb0e2268.js
+  news-web-8411e96eeb0e2268_bg.wasm
+  ```
+
+QA failure path (temporary wasm inflation, fully reverted)
+- Temporarily added a 2,000,000-byte `INFLATE` static array to
+  `crates/news-web/src/lib.rs` and referenced it from `app.rs`.
+- Rebuilt with `trunk build --release`, then ran the CI budget-check shell logic:
+  ```
+  crates/news-web/dist/news-web-9c5e74ba255bb294_bg.wasm is 2537078 bytes (budget 1800000, v1 shipped 2500000)
+  over budget by 737078 bytes
+  ```
+- Reverted both source files; `git diff` shows no changes to tracked files.
+- Post-revert rebuild produced:
+  ```
+  crates/news-web/dist/news-web-fdfd44ef4a5bb38e_bg.wasm is 537081 bytes (budget 1800000, v1 shipped 2500000)
+  ```
+
+Deviations from runway's pattern
+- Rust base image is `rust:1.89-slim-bookworm` instead of runway's 1.98, matching
+  this repo's pinned `rust-toolchain.toml`.
+- The runtime `ENV` only sets `NEWS_DATABASE_URL`; runway also sets `RUNWAY_BIND`.
+  `news-server` already binds `0.0.0.0:3000` by default, so no extra env var is
+  needed for the documented container behavior.
+- Runtime comment updated to mention the digest scheduler and quiet-hour gate
+  rather than calendar recurrence, reflecting news-triage's domain.
+
+Evidence files
+- `.omo/evidence/task-20-news-triage.txt` — this file.