Systemic TC18.txt citation-drift in src/*.rs doc comments (17 files, ~57% of checked citations) — mirrors c-RCP#434 #164

Summary

Modeled directly on c-RCP's own issue #434 ("Systemic TC18.txt citation-drift affects near-every section's .fusa-reqs.json "tc18" fields"), which found and fixed the identical root cause in a sibling repo: a TC18.txt re-extraction/rebaseline event shifted line numbers by roughly 350-450 lines across most of the document, and in places renumbered tables/figures outright.

A systematic audit of rust-RCP (mirroring c-RCP #434's own method: mechanically relocate each citation's quoted text in current TC18.txt, then hand-verify a sample) found the two citation layers are in opposite states:

• .fusa-reqs.json is essentially clean — 261 of 628 requirements carry a tc18 field, and only one genuine drift (REQ-CMP-003, ~20-24 lines) plus one completeness gap (REQ-LIN-008 missing a line number on its second quote) were found after hand-verifying ~25 flagged candidates. Both already fixed in fix: correct REQ-CMP-003/REQ-LIN-008 .fusa-reqs.json citation gaps #163. This matches evidence that an earlier [c-RCP-18-tracker Part 1]-equivalent PR series (docs(fusa): [c-RCP-18-tracker Part 1] propagate tc18_master_id for REQ-WIRE-*/REQ-NTSCF-*/REQ-TSCF-*/REQ-ABB-*/REQ-GBB-*/REQ-BMI-* #141-docs(fusa): [c-RCP-18-tracker Part 1] final scope sweep — propagate tc18_master_id for remaining prefixes, close out Part 1 #151, the tc18_master_id propagation batches) already swept this file's citations clean before this session's own ISELED/MDIO/etc. work landed.
• src/*.rs doc comments (//!//// "Provenance note" prose, not .fusa-reqs.json) were never part of that earlier sweep and are systemically drifted — the same pattern c-RCP #434 found, just in a different location within this repo.

Scope confirmed (17 of 55 source files carry TC18 citations; the other 38 have none and are clean by construction)

Of 101 lineref citations checked in doc comments, only 7 (7%) are clean; 58 (57%) show a confirmed >5-line drift, the large majority clustering at ~370-420 lines — same order of magnitude as c-RCP's own rebaseline shift. 14 of 17 affected files independently hand-verified (not just script-flagged).

• acf.rs (REQ-ABB, REQ-BMI, REQ-ECHO, REQ-EVT, REQ-GBB, REQ-RESP): worst offender — 0 clean / 13 drift / 3 not-found / 2 unpaired, drift ranging ~70 to ~400 lines. Concrete proof case: REQ-EVT-001's .fusa-reqs.json citation correctly says TC18.txt L4069-4071 for "event bits evt[2:0]..."; acf.rs's own doc comment for the exact same phrase (the Evt struct) still says TC18.txt lines 3672-3673 — off by ~399 lines. Two independently-maintained citations for the identical fact, one fixed, one not.
• avtp.rs (REQ-HVSEL, REQ-NET, REQ-NTSCF, REQ-SID, REQ-TSCF, REQ-WIRE): all 4 citations drifted (~394 lines).
• can.rs (REQ-CAN): 0 clean — 8 drift, 2 not-found, 2 unpaired (~416-line drift confirmed).
• discovery.rs (REQ-DISC): all 11 citations drifted, but by a smaller, distinct ~63-line shift — not the same ~400-line pattern, worth checking separately in case it's a different root cause.
• e2e.rs (REQ-CRC): 8 citations, all problematic — 2 confirmed ~416-line drift, plus 6 "not found" that are partly genuine RC4→RC5 content rewording (TC18.txt itself carries a Commented [BE39]: 051RC4 - text and figure updated marker at the relevant spot), not just a line-number shift — this file needs closer reading, not a mechanical line-number swap.
• gpio.rs (REQ-GPIO): 0 clean — 4 drift, 2 not-found (~408-line drift confirmed).
• i2c.rs (REQ-I2C): 1 citation, drifted ~411 lines.
• iseled.rs (REQ-ISELED): mostly script artifacts (self-referential "Provenance note" text quoting its own old, now-corrected citation) from this session's PR feat: implement Table 30 Row-2 evt[2:0] validation — ISELED (7th endpoint type) #161 fix — not new drift, but confirm before assuming fully clean.
• lifecycle.rs (REQ-LIFE): 0 clean — 3 not-found, 2 drift (~376-line drift confirmed).
• lin.rs (REQ-LIN): 0 clean — 3 drift, 1 unpaired (~416-line drift confirmed), separate from the correct LIN-specific Table-30 addition from PR feat: implement Table 30 Row-2 evt[2:0] validation — LIN (4th endpoint type) #158.
• mdio.rs (REQ-MDIO): mostly script artifacts like iseled.rs, but one genuinely still-stale citation confirmed: line 130 cites "Table 57 (TC18.txt line 5676)," directly contradicting this same file's own later correction note (lines 316-320) that the real table is Table 60 at line 6088. Even the file this session's PR feat: implement Table 30 Row-2 evt[2:0] validation — MDIO (8th and last endpoint type) #162 touched isn't fully clean.
• pwm.rs (REQ-PWM, REQ-PWMI): 0 clean — 2 drift (~408 lines).
• request.rs (REQ-BUNDLE, REQ-CANCEL, REQ-CHAIN, REQ-CMP, REQ-CRC, REQ-E2EENF, REQ-ERRH, REQ-OVRFLW, REQ-PRIO, REQ-RLC, REQ-SAFEACT, REQ-SAFEMEAS, REQ-SAFETY, REQ-SEQ, REQ-SEQENF, REQ-TIME, REQ-TRIG): 1 clean, 2 drift (~81 lines — again a smaller, distinct shift).
• timestamp.rs (REQ-TS): all 3 citations drifted (~375 lines).
• evtgroup.rs / uart.rs: mostly clean — recently-written citations from this session's own "Table 30 Row-2" PR series (feat: implement Table 30 Row-2 evt[2:0] validation — pilot (shared predicate + I2C dispatch wiring) #155-feat: implement Table 30 Row-2 evt[2:0] validation — MDIO (8th and last endpoint type) #162), not part of the original drift.

Recommended remediation

Same as c-RCP #434's own successful approach: not per-citation manual edits — a single bulk/automated correction pass, but scoped to src/*.rs doc comments specifically (the .fusa-reqs.json layer does not need this treatment; see #163).

1. For each flagged citation, re-locate its quoted spec text in current TC18.txt and replace the line range (and table/figure number, where cited).
2. Spot-check a sample against a fresh PDF extraction directly, not just TC18.txt (TC18.txt is itself one extraction of the real spec).
3. Do not assume uniform drift within a file — c-RCP #434's own closing note recorded a real instance of a claimed "whole ID-range is stale" finding turning out true for most but not all entries in that range; verify each citation individually.
4. discovery.rs's and request.rs's smaller, distinctly-sized drifts (~63 and ~81 lines respectively, vs. the dominant ~370-420-line pattern) may indicate a different or partial root cause — investigate separately rather than assuming one single shift value applies everywhere.
5. e2e.rs's "not found" citations need closer reading (possible genuine RC4→RC5 content rewording, not just a line shift) before blindly relocating them.
6. Skip iseled.rs/mdio.rs's self-referential "Provenance note" artifacts (already-fixed session work describing its own fix history) — but DO fix mdio.rs:130's genuine remaining stale citation.

Verification expectations

Doc-comment-only change (no behavior change) — full cargo build/cargo test should be unaffected; bash scripts/api-snapshot-check.sh unaffected (no public-API change); rsfusa check/rsfusa trace unaffected (citation text isn't itself trace-relevant, only the //fusa:req tags are, and those aren't touched by this fix).

References

• c-RCP issue #434 (closed) — the sibling-repo precedent this issue mirrors exactly
• fix: correct REQ-CMP-003/REQ-LIN-008 .fusa-reqs.json citation gaps #163 (this repo) — the small, already-fixed .fusa-reqs.json-layer companion to this issue
• Investigation scripts/raw results referenced in the audit that produced this issue (session-local, not committed — a fresh re-audit script should be written for the actual fix pass rather than assuming the investigation's own script is fix-ready)