Audit-only wave over the surface prior waves did not reach: src/wire.rs framing, the .github/ CI-CD workflows, the seal / .t27c-version mechanism, and the deeper src/modem.rs sample-domain paths. Report: docs/WAVE_REPORT_2026-07-11-v2.md. Draft PR follows.
Grounded against main@6850649 (verified 2026-07-11): main does NOT build (build.rs:27-31, E0308/E0599 — already filed #59/#60). 68 specs, 101 #[test], 68 gen/rust modules, 1/68 specs wired into src/, 1/68 specs sealed.
Fresh weak spots (de-duplicated against #59/#67/#69; moving-master + allowlist already being closed by in-flight #60)
| # |
Weak spot |
Sev |
File:line |
| WS1 |
Frame-sync gate is an ABSOLUTE correlation magnitude (SYNC_THRESHOLD = 8.0), not normalized to received energy. Under AD9361 AGC the Barker peak scales 13*g*A; a fixed gate never fires (low gain) or always fires (high gain). -sim-invisible; blocks first RX bring-up (#9, #11). SPS=4 is also a hard oversample lock. |
P1 |
src/modem.rs:28,74,232 |
| WS2 |
Seal / FROZEN_HASH ring is inert: .trinity/seals/ covers 1 of 68 specs and ZERO workflows verify it. Post-#60 regenerate-and-diff supersedes it, leaving a dead artifact implying a freeze CI does not enforce. |
P2 |
.trinity/seals/specs_MeshWire.json |
| WS3 |
Reproducibility is asymmetric: t27c pinned to a commit (#60) but the Rust toolchain floats (dtolnay/rust-toolchain@stable). With clippy -D warnings a new lint reddens CI on zero code change. No rust-toolchain.toml. |
P2 |
ci.yml:13,17 |
Audited clean (honest negatives): src/wire.rs framing/bounds; drift-guard spec coverage.
Prescriptions (primary refs)
- WS1: normalized sync metric — Schmidl-Cox, IEEE T-COMM 45(12), DOI 10.1109/26.650240; openwifi arXiv:2003.09525.
- WS2: verify or delete the seal — SLSA v1.0 https://slsa.dev/spec/v1.0/ ; in-toto USENIX Security 2019.
- WS3: pin the toolchain — reproducible-builds.org ; rustup toolchain-file ; OpenSSF Scorecard Pinned-Dependencies.
Full 4-sprint plan and 3 cooperation lanes in the report. This wave is audit-only by design (main does not build -> nothing verifiable end-to-end; CI surface owned by #60; WS1 is hand-written PHY logic, out of the spec-first mandate).
Anchor: phi^2 + phi^-2 = 3.
Audit-only wave over the surface prior waves did not reach:
src/wire.rsframing, the.github/CI-CD workflows, the seal /.t27c-versionmechanism, and the deepersrc/modem.rssample-domain paths. Report:docs/WAVE_REPORT_2026-07-11-v2.md. Draft PR follows.Grounded against
main@6850649(verified 2026-07-11):maindoes NOT build (build.rs:27-31, E0308/E0599 — already filed #59/#60). 68 specs, 101#[test], 68 gen/rust modules, 1/68 specs wired intosrc/, 1/68 specs sealed.Fresh weak spots (de-duplicated against #59/#67/#69; moving-master + allowlist already being closed by in-flight #60)
SYNC_THRESHOLD = 8.0), not normalized to received energy. Under AD9361 AGC the Barker peak scales13*g*A; a fixed gate never fires (low gain) or always fires (high gain).-sim-invisible; blocks first RX bring-up (#9, #11).SPS=4is also a hard oversample lock.src/modem.rs:28,74,232.trinity/seals/covers 1 of 68 specs and ZERO workflows verify it. Post-#60 regenerate-and-diff supersedes it, leaving a dead artifact implying a freeze CI does not enforce..trinity/seals/specs_MeshWire.jsont27cpinned to a commit (#60) but the Rust toolchain floats (dtolnay/rust-toolchain@stable). Withclippy -D warningsa new lint reddens CI on zero code change. Norust-toolchain.toml.ci.yml:13,17Audited clean (honest negatives):
src/wire.rsframing/bounds; drift-guard spec coverage.Prescriptions (primary refs)
Full 4-sprint plan and 3 cooperation lanes in the report. This wave is audit-only by design (main does not build -> nothing verifiable end-to-end; CI surface owned by #60; WS1 is hand-written PHY logic, out of the spec-first mandate).
Anchor: phi^2 + phi^-2 = 3.