Move logging initialization into Fastly adapter (PR 10)

[prk-Jr]

Summary

• Move Fastly logging initialization into an adapter-local module so backend setup stays owned by the Fastly adapter instead of leaking through main.rs.
• Remove log-fastly from trusted-server-core so core remains backend-agnostic and future Edgezero adapters can rely on the shared log facade only.
• Capture the PR 10 design and implementation notes used to keep this stacked migration step scoped against the PR 9 base branch.

Changes

File	Change
crates/trusted-server-adapter-fastly/src/logging.rs	Add the adapter-local Fastly logger initializer, target-label helper, and unit test.
crates/trusted-server-adapter-fastly/src/main.rs	Call logging::init_logger() and remove the inline Fastly logger implementation.
crates/trusted-server-core/Cargo.toml	Remove the log-fastly dependency from core.
Cargo.lock	Update the lockfile after removing core's log-fastly dependency edge.
docs/superpowers/specs/2026-04-02-pr10-logging-initialization-design.md	Add the PR 10 design note and scope.
docs/superpowers/plans/2026-04-02-pr10-logging-initialization.md	Add the implementation plan for the logging boundary change.

Closes

Closes #491

Test plan

• cargo test --workspace
• cargo clippy --workspace --all-targets --all-features -- -D warnings
• cargo fmt --all -- --check
• JS tests: cd crates/js/lib && npx vitest run
• JS format: cd crates/js/lib && npm run format
• Docs format: cd docs && npm run format
• WASM build: cargo build --package trusted-server-adapter-fastly --release --target wasm32-wasip1
• Manual testing via fastly compute serve
• Other:

Hardening note

Not applicable. This PR does not touch config-derived regex or pattern compilation.

Checklist

• Changes follow CLAUDE.md conventions
• No unwrap() in production code — use expect("should ...")
• Uses log macros (not println!)
• New code has tests
• No secrets or credentials committed

[aram356]

Summary

Clean, well-scoped extraction that correctly moves Fastly logging initialization into the adapter, making trusted-server-core backend-agnostic. No behavioral changes.

Non-blocking

🔧 wrench

• Unnecessary #[allow(dead_code)]: init_logger() is called from main.rs — the suppression is incorrect and would mask a real warning if the call is removed (logging.rs:8)

🤔 thinking

• _message vs message: The original used the message parameter directly (idiomatic fern). The rename to _message + record.args() is semantically identical but gratuitous (logging.rs:18)

🌱 seedling

• rsplit_once alternative: More explicit than split("::").last() for extracting the final segment (logging.rs:4)
• Missing test edge cases: No coverage for inputs without ::, empty strings, or trailing :: (logging.rs:38)

👍 praise

• Clean separation of concerns: log-fastly correctly removed from core, adapter owns its logging, pub(crate) visibility is correct
• Thorough design and plan docs: Clear scope, out-of-scope boundaries, and reproducible implementation plan

[aram356]

Summary

Clean extraction of Fastly logging initialization from trusted-server-core into trusted-server-adapter-fastly. The change is small, well-scoped, and achieves the stated goal of making core backend-agnostic for logging. CI passes.

Non-blocking

🤔 thinking

• target_label empty-result fallback: rsplit_once on "trailing::" returns "", producing log output like INFO [] message. Consider falling back to the full target when the extracted segment is empty. (crates/trusted-server-adapter-fastly/src/logging.rs:5)

📌 out of scope

• trusted-server-core still depends on fastly directly: Removing log-fastly is a step toward backend-agnostic core, but fastly = { workspace = true } remains in core's Cargo.toml. Worth tracking as a follow-up. (crates/trusted-server-core/Cargo.toml:25)

🌱 seedling

• Missing doc comments on target_label and init_logger: Both are pub(crate) and lack /// comments. init_logger is a key startup function and target_label has non-obvious semantics — a brief doc comment would help. (crates/trusted-server-adapter-fastly/src/logging.rs:4,8)

⛏ nitpick

• Plan/implementation mismatch: Plan shows split("::").last().unwrap_or(target) but implementation uses rsplit_once. Implementation is better, but the plan doc is now inaccurate. (docs/superpowers/plans/2026-04-02-pr10-logging-initialization.md:81)

CI Status

• integration tests: PASS
• browser integration tests: PASS
• prepare integration artifacts: PASS

[aram356]

Summary

Second-pass review. Prior CHANGES_REQUESTED findings from 2026-04-10 are all resolved in commit dd6929c: #[allow(dead_code)] removed, idiomatic fern message parameter restored, rsplit_once with empty-segment fallback in place, and target_label tests cover the no-separator, empty-string, and trailing-:: cases. The core ↔ adapter boundary move is clean — log-fastly is gone from trusted-server-core and lives only in trusted-server-adapter-fastly. CI is green. Approving with a few non-blocking suggestions inline.

Non-blocking

⛏ nitpick

• target_label visibility: could be private instead of pub(crate) (logging.rs:9)
• Trailing-:: fallback still contains the separator: cosmetic only; real log targets never hit this (logging.rs:13)
• Plan doc drift: plan file still references the pre-rsplit_once snippet per the prior review

🌱 seedling

• Hardcoded LevelFilter::Info: natural follow-up to wire through Settings now that logging owns its own module (logging.rs:29)
• chrono::Local::now() under WASI: effectively resolves to UTC via fallback; consider switching to chrono::Utc::now() explicitly (logging.rs:37)

👍 praise

• Clean boundary move: minimal surface, focused tests, prior feedback fully addressed (logging.rs)

CI Status

• browser integration tests: PASS
• integration tests: PASS
• prepare integration artifacts: PASS