HFDL: forensic round — LPDU emission parity, bench fixture (33 vs oracle 37)

[kevinelliott]

What

The HFDL forensic round (task #36), VDL2 methodology — dumphfdl built from source, frame-exact diff on the 21931 kHz capture:

• The 13 data LPDUs match the oracle one-for-one (every uplink/ack'd data frame xng decodes corresponds exactly).
• The missing 4 are the weakest bursts (4.0–5.0 dB SNR at 300 bps) — a genuine deep-weak sensitivity tail, same class as AIS's, not a convention bug. The VDL2-style suspicion was checked and cleared.
• Parser-policy fixes: CRC-valid LPDUs with unparsable HFNPDU contents now emit an unnumbered-data envelope (hex payload) instead of vanishing; LPDU 0x4F correctly labeled logon-resume.
• Bench gate: hfdl_offair fixture (release asset) with floor 31 (actual 33) — the fourth fenced mode.

xng now stands at 33/37 (89 %) of dumphfdl with the gap precisely characterized.

Testing

Full workspace green; all four bench gates green (this PR's CI exercises the new fixture).

🤖 Generated with Claude Code

Summary by CodeRabbit

• Bug Fixes

  • HFDL parser now distinguishes logon request and logon resume events separately.
  • Parser ensures CRC-valid unnumbered-data LPDUs emit envelope events instead of being silently dropped.

• New Features

  • Added HFDL off-air regression benchmark.

• Documentation

  • Updated benchmark documentation with HFDL benchmark details and parser improvements.

• Chores

  • Updated build infrastructure for benchmark asset downloads.
  • Updated benchmark baseline calibration data.

[coderabbitai]

📝 Walkthrough

Walkthrough

HFDL parser now emits separate events for logon request (0x8F|0xBF) and logon resume (0x4F) variants, and prevents silent drops of CRC-valid unnumbered-data LPDUs by emitting envelope events when payloads are short or unparsable. A new benchmark fixture (hfdl_48k.cs16) with baseline floor (31 frames) validates the parser improvements via CI workflow and shell script integration, documented in the notes.

Changes

HFDL Parser and Benchmark Improvements

Layer / File(s)	Summary
HFDL parser behavior improvements crates/xng-mode-hfdl/src/pdu.rs	PduParser splits logon variants into separate logon-request (0x8F|0xBF) and logon-resume (0x4F) events. HFNPDU parsing now emits unnumbered-data envelope events containing payload hex when headers don't match 0xFF or ACARS parsing fails, instead of silently returning.
HFDL benchmark fixture and baseline setup .github/workflows/bench.yml, .gitignore, bench/baselines.json, bench/run.sh	GitHub Actions workflow downloads hfdl_48k.cs16 fixture asset; .gitignore excludes it; baselines.json defines hfdl_offair minimum decode count (31); run.sh conditionally benchmarks frame counts against baseline, skipping if fixture is absent.
Benchmark documentation update docs/notes/BENCHMARKS.md	Documents HFDL event counts vs dumphfdl, explains remaining event deficit as weak-burst sensitivity tail, and describes parser-policy changes (no silent CRC-valid drops, unnumbered-data envelopes, correct logon labeling) alongside fixture and baseline updates.

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~25 minutes

Poem

A rabbit's benchmark hops with glee, ✨
Logons split true, both far and near,
No silent drops—each frame's a cheer,
From 0x4F to 0xBF's tree,
The HFDL envelope's crystal clear! 🐰📡

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title directly references the main technical objective: achieving LPDU emission parity in HFDL parsing (33 decoded frames vs 37 oracle) with the bench fixture addition, which is precisely what the changeset accomplishes.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

📝 Generate docstrings

• Create stacked PR
• Commit on current branch

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch hfdl-lpdu-parity

---

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 and usage tips.}

[coderabbitai]

🧹 Nitpick comments (1)

crates/xng-mode-hfdl/src/pdu.rs (1)

198-229: ⚡ Quick win

Add focused unit tests for the new fallback/classification branches.

Please add parser tests that assert: (1) LPDU 0x4F emits logon-resume, and (2) CRC-valid but unparsable HFNPDU content emits unnumbered-data with stable data_hex. The bench decode-count gate validates totals, but not event-kind/detail correctness.

Proposed test scaffold

+    #[test]
+    fn lpdu_0x4f_emits_logon_resume() {
+        let lpdu = with_fcs(vec![0x4F, 0x12, 0x34, 0x56]);
+        let mpdu = build_mpdu_downlink(3, 0xC7, &[lpdu]);
+        let ev = PduParser::new().parse(&mpdu, 300);
+        assert_eq!(ev.len(), 1);
+        assert_eq!(ev[0].kind, "logon-resume");
+    }
+
+    #[test]
+    fn unparsable_hfnpdu_emits_unnumbered_data_envelope() {
+        let lpdu = build_lpdu_hfnpdu(&[0x01, 0x02]); // non-0xFF HFNPDU
+        let mpdu = build_mpdu_downlink(3, 0xC7, &[lpdu]);
+        let ev = PduParser::new().parse(&mpdu, 300);
+        assert_eq!(ev.len(), 1);
+        assert_eq!(ev[0].kind, "unnumbered-data");
+        assert_eq!(ev[0].details["data_hex"], "0102");
+    }

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@crates/xng-mode-hfdl/src/pdu.rs` around lines 198 - 229, Add focused unit
tests for the HFDP/LPDU parser to cover the new fallback/classification
branches: create tests that feed a LPDU with first byte 0x4F and assert the
parser emits an HfdlEvent with kind "logon-resume" and appropriate details; and
feed a CRC-valid HFNPDU whose payload bytes are not parsable by
xng_acars::block::parse and assert the parser emits an HfdlEvent with kind
"unnumbered-data" whose details contain a stable data_hex exactly matching the
input bytes and raw equals the original payload. Locate the parsing entry (in
pdu.rs where the envelope closure, match on h[1], and xng_acars::block::parse
are defined) and add tests that call the public parse function/entrypoint used
by the crate, constructing minimal headers/CRC so the code path hits the 0x4F
and the CRC-valid but unparsable branches, and assert on HfdlEvent.kind,
details["data_hex"], and raw fields.

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Nitpick comments:
In `@crates/xng-mode-hfdl/src/pdu.rs`:
- Around line 198-229: Add focused unit tests for the HFDP/LPDU parser to cover
the new fallback/classification branches: create tests that feed a LPDU with
first byte 0x4F and assert the parser emits an HfdlEvent with kind
"logon-resume" and appropriate details; and feed a CRC-valid HFNPDU whose
payload bytes are not parsable by xng_acars::block::parse and assert the parser
emits an HfdlEvent with kind "unnumbered-data" whose details contain a stable
data_hex exactly matching the input bytes and raw equals the original payload.
Locate the parsing entry (in pdu.rs where the envelope closure, match on h[1],
and xng_acars::block::parse are defined) and add tests that call the public
parse function/entrypoint used by the crate, constructing minimal headers/CRC so
the code path hits the 0x4F and the CRC-valid but unparsable branches, and
assert on HfdlEvent.kind, details["data_hex"], and raw fields.

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro

Run ID: f7b16ea5-cab0-4acc-b4e2-0e099756f0ea

📥 Commits

Reviewing files that changed from the base of the PR and between fe03312 and 5d7b4e8.

📒 Files selected for processing (6)

• .github/workflows/bench.yml
• .gitignore
• bench/baselines.json
• bench/run.sh
• crates/xng-mode-hfdl/src/pdu.rs
• docs/notes/BENCHMARKS.md