The ATL core library’s test suite passes at a pinned revision.

atl-core tests exit 0 at a pinned revision.

Used in: Tamper-Evident Audit Logs, RFC 3161 vs OpenTimestamps, AI Agent Audit Trail.

Watch

Needs: a browser. Not timed.

CI: run CI, 2026-09-25 , executed by GitHub Actions, not by me.

You should see: jobs Format, Test, MSRV (1.92.0), Clippy, Coverage, all green.

Would falsify: a failed job.

You should see: eight test result: ok. N passed; 0 failed; 0 ignored; 0 measured; 0 filtered out lines (the same 857 tests, 0 failed, as the Run section), then benches/merkle_bench.rs’s per-function checks ending Testing format_signature then Success.

Would falsify: a FAILED test-result line, a non-zero exit, or a bench-function line reporting anything other than Success.

Typed replay of the verbatim local transcript of 2026-09-25 at pin 6e652ed, run in a clean git worktree: every command shown is the literal command run, every output line is the literal output. One thing is dropped, not rewritten: cargo’s own dependency Compiling ... chatter (145 lines, from a worktree with no cached target/) — none of it repo-specific, replaced by a bracketed note; every test-result line is untouched. Output timing is not execution timing. The viewer did not reproduce it.

Run

Needs: Rust 1.92 or newer. The CI Test job took 1 min on a GitHub Actions runner.


git clone https://github.com/evidentum-io/atl-core.git && cd atl-core && git checkout 6e652ed
cargo test --all-targets
# exit 0

You should see: eight test result: ok. N passed; 0 failed; 0 ignored; 0 measured; 0 filtered out lines — unit tests plus arbitrary_precision_audit, integration, lib_exports_tests, ots_calendar_integration, ots_fixtures, proptests, and rfc_vectors (731 + 5 + 28 + 33 + 6 + 6 + 26 + 22 = 857 tests, 0 failed anywhere), then benches/merkle_bench.rs’s per-function Success lines, exit 0.

Would falsify: a non-zero exit from cargo test --all-targets, or any suite reporting a failure.

Read

Needs: nothing to install.

RFC 8785 canonicalization is canonicalize and check_unique_property_names in jcs.rs: numbers are rendered with ryu-js’s ECMAScript algorithm (Rust’s own f64 Display is not RFC 8785 §3.2.2.3-conformant), and duplicate property names are refused (check_unique_property_names) rather than silently resolved, since neither RFC 8785 nor RFC 8259 §4 says which occurrence should win. verify_receipt_anchors in facts.rs is the per-anchor fact orchestration atl-cli’s own verdict layer builds on: every non-passing check is recorded as refuted (something checkable is false) or indeterminate (nothing was refuted, but a check could not be finished) — never collapsed to a single boolean. Full RFC 8785 rationale and the anchor-chain trust distinctions in README.md .

Test suites: eight cargo test --all-targets binaries (unit tests plus arbitrary_precision_audit, integration, lib_exports_tests, ots_calendar_integration, ots_fixtures, proptests, and rfc_vectors), 857 tests total. rfc_vectors.rs checks against frozen, externally-sourced vectors (Certificate Transparency Merkle values, RFC 6962 algorithm text, RFC 8785 JCS values, and ATL-protocol-derived values) that the crate never computed itself; src/core/verify/tests/rfc3161_corpus_tests.rs checks a corpus of 16 real RFC 3161 timestamp tokens issued by four independent free TSAs (FreeTSA, Sectigo, DigiCert, GlobalSign); proptests.rs property-tests JCS canonicalization and Merkle root/inclusion/consistency proofs.

Does not establish

coverage
an oracle
that the suite is complete

If this does not survive your check, write me.