feat(wire): T27-first partial flip — specs/wire.t27 becomes SSOT by gHashTag · Pull Request #33 · gHashTag/tri-net

gHashTag · 2026-07-04T11:02:23Z

Summary

Flips src/wire.rs to consume auto-generated code from specs/wire.t27 via t27c gen-rust. The spec is now the source of truth for constants and pure predicates; the Rust module only adds ergonomic wrappers on top.

What changed

gen/rust/wire.rs (new, deterministic build output): auto-generated constants + predicates, plus documented hand-written stubs for be_byte / u32_be (see limitation below).
src/wire.rs: re-exports auto-gen symbols, keeps Header / FrameKind / to_bytes / parse on top. No local re-declaration of VERSION or HEADER_LEN anymore.
build.rs (new): optional t27c gen-rust invocation under T27C_REGENERATE=1. Silent on missing t27c, so CI without the compiler still builds.
docs/T27_PORT_STATUS.md: wire.rs row updated to 'T27-FIRST (partial)'.
docs/T27_FIRST_MIGRATION.md (new): rationale, bootstrap limitation, regeneration recipe.

Why partial

t27c-0.1.0 bootstrap has a parser bug on bit-shift expressions — both gen-rust and gen (Zig) drop ((w >> N) & 255) as u8 and emit return ();. So be_byte / u32_be cannot come from the compiler yet. They live inside gen/rust/wire.rs beneath a documented banner. Once the bit-shift bug is fixed upstream in gHashTag/t27, delete the stubs and let gen-rust emit them. Filing a repro issue against gHashTag/t27 as a follow-up.

Verification

cargo test --lib — 101/0 (includes two new guardrails: constants + predicates).
cargo test --test m2_routing_pure_logic — 25/0.
cargo fmt --all -- --check — clean.

No hardware runs, no fabricated metrics, no changes to crypto / modem / routing.

Anchor: phi^2 + phi^-2 = 3

Constants (VERSION, KIND_HELLO, KIND_DATA, HEADER_LEN) and pure predicates (frame_kind_valid, header_byte, parse_accepts) are now auto-generated into gen/rust/wire.rs via t27c gen-rust. src/wire.rs re-exports them and keeps only the ergonomic Rust wrappers (Header struct, FrameKind enum, to_bytes, parse) on top. Partial rather than full flip: t27c-0.1.0 has a bit-shift parser bug that emits 'return ();' for expressions like ((w >> 24) & 255) as u8, so be_byte and u32_be stay hand-written inside gen/rust/wire.rs (below a documented banner) instead of coming from the compiler. Bug reproduces on gen-rust and gen (Zig) both — it is in the shared frontend, not the emitter. To be filed against gHashTag/t27. build.rs invokes t27c only when T27C_REGENERATE=1 is set, so CI without t27c installed still builds. gen/rust/ is committed as a deterministic build output. Guardrails added: - t27_gen_constants_match_hand_written — pins auto-gen constant values. - t27_gen_predicates_match_semantics — exercises the generated predicates. Verified green: - cargo test --lib: 101/0 - cargo test --test m2_routing_pure_logic: 25/0 - cargo fmt --all -- --check: clean Docs: - docs/T27_PORT_STATUS.md — wire.rs row flipped to 'T27-FIRST (partial)'. - docs/T27_FIRST_MIGRATION.md — new; documents the flip, the bootstrap limitation, the build story, and the regeneration recipe. Anchor: phi^2 + phi^-2 = 3

…hift Post-merge triage on the t27c bootstrap isolated the real trigger. The bug is not in bit-shift parsing; it is missing ExprCast lowering in the Rust, Zig, and C text emitters. The frontend builds the AST node correctly, but expr_to_rust and its Zig/C peers have no arm for NodeKind::ExprCast and fall through their default arm to the unit value '()'. gen-verilog is the only backend that implements ExprCast. Isolation table (all against current master t27c): w >> 24 -> OK (w >> 24) & 255 -> OK (w >> 24) as u8 -> () <-- cast triggers it (w & 255) as u8 -> () <-- no shift, same failure 42 as u32 -> () x as u32 -> () ((a as u32) << 8) | ... -> ((() << 8) | ()) Filed as gHashTag/t27#1314 with the isolation matrix and the compiler.rs line reference. Practical impact on this PR: unchanged. be_byte and u32_be still cannot come from t27c because both use 'as u8' / 'as u32'. Hand-written stubs stay beneath the banner in gen/rust/wire.rs until t27c ships an ExprCast arm in expr_to_rust. This commit only fixes the wording in the banner and the migration doc so the tracking issue is accurate. Anchor: phi^2 + phi^-2 = 3

gHashTag · 2026-07-04T11:14:47Z

Post-merge triage note before human-merge: upstream root cause is precisely identified and filed as gHashTag/t27#1314. The bug is missing ExprCast lowering in the Rust/Zig/C text emitters, not bit-shift parsing (my earlier guess in-thread). Isolation matrix and the exact compiler.rs fallthrough point (expr_to_rust at line 8054, default arm at line 8172) live in the issue.

Follow-up commit 880954e updates the banner in gen/rust/wire.rs and docs/T27_FIRST_MIGRATION.md so the tracking reference is honest. No code paths changed; still green:

cargo test --lib: 101/0
cargo test --test m2_routing_pure_logic: 25/0
cargo fmt --all -- --check: clean

Once t27c#1314 lands, follow-up PR: strip the hand-stub band from gen/rust/wire.rs, regenerate via T27C_REGENERATE=1 cargo build, verify diff.

Anchor: phi^2 + phi^-2 = 3

gHashTag · 2026-07-04T12:07:08Z

Upstream unblock landed: gHashTag/t27#1320 — ExprCast arm added to expr_to_rust (before default), mirrors the gen-verilog arm. t27c self-tests 20/0, gen-rust on specs/wire.t27 now lowers be_byte and u32_be correctly (was return ();).

Merge order once approved:

Human-merge t27#1320 into t27:master.
Rebuild t27c from that master.
In tri-net feat/t27-first-wire: T27C_REGENERATE=1 t27c gen-rust specs/wire.t27 > gen/rust/wire.rs, verify diff drops the hand-stub band only, re-run tests.
Then human-merge this PR feat(wire): T27-first partial flip — specs/wire.t27 becomes SSOT #33.

Zig + C emitters remain follow-up under t27#1314 (Rust arm in #1320 is the template). tri-net does not consume them today so this does not block the flip.

Anchor: phi^2 + phi^-2 = 3

…20 + gen/ untouchable rule - Two remaining 'bit-shift' references (build.rs note, Next flips) corrected to ExprCast lowering (t27#1314 → fixed by t27#1320). The Bootstrap-limitation section was already corrected in 880954e; these two were missed. - Regeneration recipe updated to reflect pre-validated reality: post-#1320 regen overwrites the whole gen file (~49 removed / ~40 added), NOT just the stub band. Tests stay 101/0 + 25/0. Bigger diff is expected/correct (removes obsolete ExprCast banner + hand-stubs, normalises cosmetic rendering). - New 'Rule: gen/ is untouchable raw output' section: gen/<lang>/* is never hand-edited; explanatory banners/comments go in the consumer (src/) or this doc. If t27c emits wrong, fix t27c upstream, don't patch gen/. Captures the diff-shape lesson so future flips (routing, gf16) don't repeat the trap.

…f '()' (Closes #1314) Add Expr::Cast arm in expr_to_rust (bootstrap/src/compiler.rs), mirroring the gen-verilog arm at compiler.rs:4941. Before this fix, ExprCast fell through to the default '_ => "()".to_string()' branch and generated empty-tuple stubs in Rust output, silently corrupting any T27 spec using bit-width casts. Also update docs/NOW.md with an exprcast-rust-emitter entry (required by the NOW Sync Gate). Closes #1314 (gen-rust half; gen-c and gen-zig follow-ups tracked separately). Unblocks gHashTag/tri-net#33 T27-first wire flip. Anchor: phi^2 + phi^-2 = 3

…f '()' (Closes #1314) (#1320) Add Expr::Cast arm in expr_to_rust (bootstrap/src/compiler.rs), mirroring the gen-verilog arm at compiler.rs:4941. Before this fix, ExprCast fell through to the default '_ => "()".to_string()' branch and generated empty-tuple stubs in Rust output, silently corrupting any T27 spec using bit-width casts. Also update docs/NOW.md with an exprcast-rust-emitter entry (required by the NOW Sync Gate). Closes #1314 (gen-rust half; gen-c and gen-zig follow-ups tracked separately). Unblocks gHashTag/tri-net#33 T27-first wire flip. Anchor: phi^2 + phi^-2 = 3 Co-authored-by: Perplexity Computer <agent@perplexity.ai>

Regenerated via t27c (release, built from t27 master c4dc8ee) which now includes the Expr::Cast arm in expr_to_rust that emits '(operand as target)' instead of the empty-tuple stub. This closes the T27-first partial flip that a6bb0b0 opened: gen/rust/wire.rs is now the full canonical output of specs/wire.t27, not a hand-patched shim. Diff shape: +40 / -49 (banner shrinks to the raw t27c header, be_byte / u32_be / header_byte bodies swap their return-() stubs for real 'as u8' / 'as u32' casts). Tests: cargo test --lib 101/0, cargo test --test m2_routing_pure_logic 25/0. Anchor: phi^2 + phi^-2 = 3 Refs #33

Real t27c gen-rust emits idiomatic-T27 output (literal return + explicit parens) which fails 'cargo clippy -- -D warnings' with needless_return, unnecessary_parens, and unused_parens. Since gen/ is untouchable (see docs/T27_FIRST_MIGRATION.md), scope the allow to the wrapping mod in src/wire.rs rather than hand-editing the generated file. Cleaner Rust rendering (drop trailing 'return', drop wrapping parens on whole-expression returns and on if-condition wrappers) is upstream work on gHashTag/t27 in expr_to_rust; when that lands, the allow can shrink or disappear on the next regeneration. cargo build 101/0 lib, m2_routing_pure_logic 25/0, cargo fmt clean. Anchor: phi^2 + phi^-2 = 3 Refs #33

Replace `u32::from_be_bytes(b[2..6].try_into().ok()?)` (and the dst-slot equivalent) with `u32_be(b[2], b[3], b[4], b[5])` — the auto-generated function produced by t27c from specs/wire.t27. Byte-order equivalence: - `u32::from_be_bytes([b0,b1,b2,b3])` = (b0<<24)|(b1<<16)|(b2<<8)|b3. - t27c-generated `u32_be(b0,b1,b2,b3)` = same expression (see gen/rust/wire.rs). - `try_into().ok()?` on a 4-byte slice never returns None post the length check on line 82; removing it eliminates a dead error path. Consequence: the parse-side arithmetic now lives under the SSOT contract. spec-drift-guard CI (workflow spec-drift-guard.yml, merged in #35) is the enforcement mechanism — any drift between specs/wire.t27 and gen/rust/wire.rs now covers this reassembly path as well. Closes weak-spot audit finding #5 (post-#33 loop): src/wire.rs still used std::from_be_bytes rather than delegating to the auto-gen equivalent, so the parse-path was outside the SSOT umbrella even though the serialize-path (header_byte + be_byte) was inside. Also re-exports `be_byte` and `u32_be` from the `gen` module for symmetry with the other spec-driven helpers. Verification (local): - `cargo build`: clean. - `cargo test --lib`: 101 passed / 0 failed. In particular `wire::tests::header_roundtrips` still passes, confirming byte-order equivalence end-to-end. Anchor: phi^2 + phi^-2 = 3. Co-authored-by: Perplexity Computer <agent@perplexity.ai>

…dit-trail primitive Bounded expansion (not full v1). The skeleton previously proposed the auditability primitive abstractly (Sections 1-4); this commit adds Section 5 reporting that the primitive is now empirically realized end-to-end at tri-net main dc1bebb + t27 master 3c912d9. Section 5 subsections: - 5.1 What is materialized -- one spec (specs/wire.t27), three generated backends (gen/{rust,zig,c}/wire.*), one workflow (spec-drift-guard.yml), consumer path under the same umbrella (src/wire.rs -> u32_be). - 5.2 Audit-trail tuple (concrete instance) -- (tri-net@dc1bebb, t27@3c912d9, spec-drift-guard run) is the minimal fetch set for a third party to independently verify byte-identity. - 5.3 Merge chain table -- 5 tri-net PRs (#33/#34/#35/#37/#38) + 2 t27 PRs (#1320/#1337) with merge SHAs. - 5.4 Empirical checks at HEAD -- three diff -u results (all empty), zero 'unsupported: ExprCast' in Zig/C output, 101 lib tests green including header_roundtrips as empirical byte-order-equivalence proof. - 5.5 What this reference impl does NOT show -- byte-identity is not functional correctness; only one spec is covered so far; no silicon; trust in t27c itself is not eliminated (moved, per Carrone 2026). Sync updates across the doc: - Header cross-link: 91a5b63 -> dc1bebb (post-#38). - Section 3.1: describes all three gen targets, not just Rust. - Section 3.2: drift-guard covers Rust + Zig + C; links to #35 + #38. - Section 3.3: ExprCast marked resolved on all four backends (Rust #1320, Zig+C #1337, Verilog pre-#1320). - Section 6 (Limits): removed obsolete 'Rust-only SSOT until #1333' bullet; added trust-surface-of-t27c and no-HDL-target-yet bullets. - Section 7 (Future work): struck through completed Zig+C drift-guard item; added protocol-stack expansion, t27c deterministic-build items. - Appendix A: three-target reproducibility recipe (8 steps), pins tri-net@dc1bebb + t27@3c912d9. Draft only -- not for external circulation. Anchor: phi^2 + phi^-2 = 3.

@intcast

Second + third T27-first flips (after wire.t27 #33). Both specs produce CLEAN gen-rust output (0 broken casts, 0 unsupported) — no dynamic arrays needed, ExprCast lowering (t27@3c912d9) handles all 'as u8'/'as u32'. Generated (6 files, 3 targets × 2 specs): gen/rust/hello.rs (67L) — MAX_HEARD, hello_byte, reports_hearing, etc. gen/rust/etx.rs (68L) — OPTIMISTIC, DEAD_EPS, ONE_FP, calc_etx, etc. gen/zig/hello.zig — @as/@intcast form gen/zig/etx.zig gen/c/hello.c — ((uint8_t)(...)) form gen/c/etx.c Guardrail tests (tests/t27_gen_hello_guardrail.rs, 4 tests): - hello constants (MAX_HEARD=3, HEADER_LEN=13) - hello byte-layout (src field BE, n at idx 8) - etx constants (OPTIMISTIC, DEAD_EPS, ONE_FP in Q8.8) Gate: cargo test --all = 141 passed, 0 failed (was 137 + 4 new). Drift-guard extension (to cover hello + etx) follows in next commit.

…-end measurement Single-process loopback confounded every metric and repeatedly produced broken-ruler results: sent and recv were the SAME node's counters (delivery ratios >100% from timing offsets, waves #26/#33/#37), jitter==0 meant "no stream" not "clean link" (#24), and a cross-thread crash-race hid until it fired (#33). There was no way to verify the UI or a real MITM either. Fix: run TWO independent TriNetMonitor instances that dial each other over real UDP. Three dev-only env hooks (no-ops in a shipping run): TRINET_LOG=<path> per-instance log (LogBus) so each process's counters read apart TRINET_LISTEN=<port> this instance's UDP listen port TRINET_AUTOCALL=host:port auto-dial a 1-1 call on launch (INVITE bypassed) CallManager.startCall now honors a distinct listen port so two locals don't collide. smoke/two_endpoint_rig.sh orchestrates A(:8000) <-> B(:8100) and reports the CROSS-process delivery A->B = (B received)/(A sent) — the honest number a single process cannot give. Verified live: two real windowed processes establish a bidirectional call (macOS shares the camera, so both send AND receive). Clean link: A->B 98.7%, B->A 93.4%. With 25% induced packet loss on B: A->B still 96.7% -- i.e. the layered loss-recovery stack (grouped FEC + whole-NAL NACK + per-fragment NACK from #31-33) recovers 25% loss to ~97% frame delivery, now proven over two genuine processes instead of a self-loopback. Mac-only (the rig runs two Mac instances); the env hooks are inert dev tooling. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

gHashTag added mesh TRI-NET mesh track documentation Docs labels Jul 4, 2026

gHashTag mentioned this pull request Jul 4, 2026

fix(gen-rust): lower ExprCast (emit 'as Type' instead of '()') gHashTag/t27#1320

Merged

gHashTag force-pushed the feat/t27-first-wire branch from 7f0db07 to 45d84ce Compare July 4, 2026 13:02

gHashTag marked this pull request as ready for review July 4, 2026 13:05

gHashTag merged commit 77a9a49 into main Jul 4, 2026
2 checks passed

gHashTag deleted the feat/t27-first-wire branch July 4, 2026 13:05

This was referenced Jul 4, 2026

ci(spec-drift-guard): enforce specs/wire.t27 -> gen/rust/wire.rs SSOT byte-for-byte #35

Merged

gen-zig and gen-c: ExprCast falls through default arm (silent drop / '/* unsupported */') gHashTag/t27#1333

Closed

gHashTag mentioned this pull request Jul 4, 2026

docs: Competitor Matrix — 10 MANET vendors with prices (2026-07-04) #28

Closed

Navigation Menu

feat(wire): T27-first partial flip — specs/wire.t27 becomes SSOT - #33

feat(wire): T27-first partial flip — specs/wire.t27 becomes SSOT#33
gHashTag merged 5 commits into
mainfrom
feat/t27-first-wire

gHashTag commented Jul 4, 2026

Uh oh!

gHashTag commented Jul 4, 2026

Uh oh!

gHashTag commented Jul 4, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Conversation

gHashTag commented Jul 4, 2026

Summary

What changed

Why partial

Verification

Uh oh!

gHashTag commented Jul 4, 2026

Uh oh!

gHashTag commented Jul 4, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant