[DecisionNerd]

Why

graphforge-cli measured 79.34% against its 80% per-crate floor and was the only crate below floor. The largest coherent uncovered block was the gf query command wrapper: the underlying execute_to_result_sink_with_evidence contract is covered at the API level, but the wrapper that selects sink options, maps the format, and renders the two receipt shapes was not exercised at all.

Coverage

Measured with the same production filter coverage_rust_ledger.py applies, so these are production lines and not raw LCOV counts.

	Covered / measured	Percent
Before	2,853 / 3,596	79.34%
After	2,902 / 3,596	80.70%

49 previously uncovered production lines, 44 in portable_cli.rs and 5 in the crate root dispatch. The floor needed 24, so this clears it with margin rather than scraping past.

Tests

Success path, both receipt forms, a boundary, and a real failure:

• the JSON receipt reports the sink contract, exact row and byte counts, and a result fingerprint;
• the human receipt reports the same outcome in its own format;
• --format arrow-ipc selects the non-default sink;
• explicit batch and row-group bounds change how the sink is written without changing the result fingerprint;
• a query the binder rejects fails, writes no partial sink, and emits a diagnostic.

Mutation evidence

cargo mutants --file crates/graphforge-cli/src/portable_cli.rs --package graphforge-cli

22 mutants: 14 caught, 3 missed, 5 unviable. The run_query whole-body stub is caught, so the tests fail when the function stops doing its work.

All three survivors are the same function, oci_credential at line 208, which this tranche does not target. It reads GRAPHFORGE_OCI_CREDENTIAL and feeds the OCI publish and pull paths. Killing them needs either a registry or process-global environment mutation, which is unsafe under parallel test execution in current Rust. The behaviour they guard is worth a follow-up, because a mutant that fabricates a credential rather than reading the environment currently goes undetected. Justified individually rather than waived in bulk, per #360.

Also

Adopts cargo-mutants in docs/engineering/TESTING.md as a developer tool, with the bounded per-module invocation. It is installed, not depended on, and adds nothing to the build.

Note for anyone running these locally: project tests need TMPDIR on a filesystem the admission policy accepts. The default temporary directory is refused as filesystem_class_unproven.

Refs #360

🤖 Generated with Claude Code

---

^{Need help on this PR? Tag @codesmith-bot with what you need. Autofix is disabled.}

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on this repository. Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Advanced

Run ID: e8ef5186-a78e-4858-84a6-aab1bbf13601

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

Warning

Billing warning: we have not been able to collect payment for this subscription for more than 72 hours. Please update the payment method or pay any pending invoices in Billing to avoid service interruption.

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}