Commit Graph
6 Commits
Author SHA1 Message Date
glm-5.3-flash 048493fefa producer: the BIND phase model + the socks5:bind command gate (ADR-006/ADR-007)
The TCP BIND dispatch arm replaces its refusal with the full phase
model (POC #5-validated scope), plus the command gate at dispatch:

- The command gate: TCP BIND requires socks5:bind from the per-call
  opener identity (CF-005/CF-006) — the pump handler resolves it
  from the open-op AuthContext; serve_socks5_over<T> takes the
  resolved can_bind boolean explicitly. A scope-less identity gets
  the RFC-faithful ConnectionNotAllowed (0x02) with no factory call
  (the gate precedes the seam) while CONNECT still serves it.
- The phase model (wire.md §BIND): the BindAcceptFactory at the
  command-read interception point → reply#1 carrying BND.ADDR (the
  factory's bind_addr) → the quiet wait (single-shot select!, exactly
  two exits: the one accept, or the control stream's EOF/garbage/
  error; any byte = protocol garbage → session ends, no data plane;
  the accept future's drop IS the client-gone teardown) → reply#2
  carrying the peer address (via the fork's published new_reply) →
  pump_bidi inline (the CONNECT shape, R-02).
- Two-stage RFC-faithful failure mapping: factory failure → reply#1's
  mapped code; accept-stage failure → reply#2's mapped code, no data
  plane. No phantom listeners, no phantom data planes.
- The no-factory default is the wrapper-aware in-band variant
  (ADR-007 variant 3, documented): sentinel reply#1 + sentinel reply#2
  (0.0.0.0:0), the full phase model preserved — the channel is the
  connection.
- The accept wait lives in the teardown-governed pump phase, never
  under the establishment bound (ADR-007).

Tests: 16 unit tests (raw RFC client over duplex) + 7 harness
integration tests over channels — happy path, the gate (0x02 while
CONNECT serves, both scopes), quiet-wait garbage, client-gone
teardown (owner-flag witness), both failure stages over channels,
no-factory variant 3, and the three-command regression guard (CONNECT
serves, BIND serves under its gate, ASSOCIATE still refuses).

Verification: cargo test green (96 lib + 26 harness + 37
differential); clippy -D warnings clean (all targets, native +
wasm32-unknown-unknown); cargo fmt clean; wasm check clean.

Task: tasks/implementation/core-producer-bind.md (status → completed)
2026-09-23 19:58:31 +00:00
glm-5.3-flash cfacc8a968 producer: the session-gate establisher + the command-dispatch pump + CONNECT
The crate's core task (implementation/core-producer-connect): the
producer half of src/producer.rs per producer.md, adapted to the
crate's seam types and the alkcall 0.8.0 establishment surface.

- The session-gate establisher (`socks5_session_establisher`):
  validates/records the opener identity (the CF-005/CF-006 seam,
  ADR-005) via the optional `identity_witness` probe (the alktunnels
  posture); returns `Establishment::default()` — no plan payload
  (the dial happens post-establishment; POC #1's wire note).
- The pump handler (`make_socks5_pump_handler`): `accept_bi()` →
  the forked typestate over the `BiStream` (`accept_no_auth →
  read_command`) → dispatch. CONNECT: the dial seam (`DialFn`,
  ADR-004) at the command-read interception point; success:
  `reply_success(sentinel)` then `pump_bidi` awaited inline (R-02 —
  the JoinHandle tracks the data-plane lifetime). Refusal:
  `reply_error` with the mapped RFC code — no phantom data plane.
- The ASSOCIATE/BIND arms land as RFC-faithful refusals
  (`CommandNotSupported`) — the pre-vertical posture (ADR-008); the
  dispatch skeleton + registration signature carry the seam
  parameters from day one so the associate/bind tasks swap the arms
  without touching the registration shape.
- `register_socks5_openable`: spec + establisher + pump via
  `register_openable_with_establisher` (`timeout: None` = the 10s
  default), scope-gated `socks5:open`. Missing callbacks = those
  commands refuse RFC-faithfully, never a panic.
- `serve_socks5_over<T>`: the substrate-free protocol entry (the
  POC shape), directly testable without channels; the builder
  (`Socks5Producer::builder()`) is the spec'd assembly shape.
- Tests: 11 in-module (substrate-free duplex-driven — round-trip +
  half-close teardown, refusal mapping per variant, missing-dial
  refusal, ASSOCIATE/BIND arms, dispatch guard, establisher witness,
  builder shape) + 5 harness integration (end-to-end CONNECT echo
  with the identity witness, dial refusal 0x05 with no phantom data
  plane, scope-class refusal 0x02, ASSOCIATE/BIND refusals with
  CONNECT still serving, repeated CONNECTs on one registration).
  The harness retargeted to the real producer (`RegisterMode::
  Producer`); the minimal stand-in is a harness-internal helper.

Verification: cargo test 87+19+37 green; clippy --all-targets -D
warnings clean; fmt clean; wasm check + wasm clippy clean; cargo
doc clean; --all-features green.
2026-09-23 19:39:08 +00:00
glm-5.3-flash f8773521b9 harness: the producer/consumer integration test harness over duplex
tests/harness.rs — the producer<->consumer wiring over one
tokio::io::duplex carrying the channels 8-byte chunk format, the
alkcall 0.8.0 establishment surface consumed for the first time by
this crate (register_openable_with_establisher + the call/adopt
dance) — the first-consumption risk retired.

- Identities: consumer (socks5:open) / bind (socks5:bind only) /
  both-scopes (explicit, ADR-006) + TestIdProvider (token path).
- In-process seams, no sockets, no `local`: echo_dial /
  refusing_dial (variant-preserving), echo_egress (readiness
  contract verified — next_reply parks Pending when idle) /
  refusing_egress, echo_accept / refusing_accept_factory /
  refusing_accept (two-stage BIND refusal shapes, mapped through
  to_reply_error()).
- Producer-side registration helper: register_socks5_openable_minimal
  + socks5_session_establisher + RegisterMode — the single place the
  producer wiring lives; the CF-006 identity witness rides it and is
  asserted end-to-end.
- RawRfcClient: the hand-rolled RFC 1928 client helper over any
  AsyncRead + AsyncWrite, round-tripped against a byte-level
  hand-rolled server (success + refusal replies).
- Smoke test: harness up -> producer registered -> consumer open +
  adopt -> raw RFC CONNECT -> echo round-trip through pump_bidi.
  Fail-closed probe: scope-less identity -> FORBIDDEN, no channel
  anywhere (the load-bearing POC #1 set_identity wiring requirement,
  explicit in the install hook and tested).

14 harness tests. Verification: cargo test (127 green), clippy
(all-targets + wasm32) clean, fmt clean, doc zero warnings,
test --all-features green, both wasm checks clean (harness is
dev-only).
2026-09-23 18:48:16 +00:00
glm-5.3-flash 7bd5d6a7e7 review 001 remediation: the forked core's findings resolved
Batch 1 (F-1 + N-1 + S-2 + N-4):
- reply_success/reply_error write replies with write_all (partial-write
  hazard closed; the err_when-chaining idiom suppresses
  clippy::unused_io_amount — reproduced fresh via a five-variant probe)
- rfc1928.md's write_all bullet records both conversions and the real
  rule (replies are single-shot, all-or-nothing) + the lint blind spot
- reject's doc names the 0xff vocabulary choice (RFC 1929 needs only
  non-0x00); Addr::Domain's deliberating comment pinned

Batch 2 (coverage: L-1 + L-2 + L-3 + S-1):
- client v6 request encoding pinned (22 bytes, ATYP=4, slice bounds)
- Socks5Stream::shutdown() EOF/BrokenPipe propagation pinned
- wait_on_tcp's io-error arm + SocksServerError::to_reply_error's full
  mapping table + read_command's command-phase version mismatch covered
- S-1: read_address's 1-byte-field bound documented; decode-side
  255/256 boundary pinned both faces

Batch 3 (posture notes: C-1, C-2, N-2, N-3 — decided keep-with-note):
- ConnectError: hand-port map's companion (cut if the producer layer
  lands without consuming it); StandardAuthentication: ADR-005
  auth-split is the anticipated consumer; CheckResult same shape
- tracing: the fork's declared logging seam, noted in Cargo.toml

Coverage 92.63% → 93.09% regions / 95.40% → 96.07% lines; the named
gaps closed, the remainder is accepted-uncovered with disposition.
Review doc carries the re-verification + resolution record.

Verification: cargo test (39 lib + 37 differential), clippy
(all-targets + wasm32) -D warnings, fmt, doc, wasm check, MSRV 1.88,
publish dry-run — all green.
2026-09-23 06:25:27 +00:00
glm-5.3-flash f8c34444fb rfc1928: the targeted fork — rename + convention pass (ADR-013)
The mechanical half of the posture change: src/vendor/ ->
src/rfc1928/ (module alksocks::rfc1928), tests/vendor-differential.rs
-> tests/rfc1928-differential.rs (git mv — history preserved).

- Fork point pinned: the tree as of 5b18d4d (the extraction review's
  gate commit); provenance permanent in the module docs and the
  extended LICENSE notice (upstream MIT preserved, fork-point +
  post-fork modification statement added — extend, never replace)
- The lint-allow posture dissolved: all module-level #[allow]s
  removed and the suppressed smells fixed (u8::max_value() ->
  u8::MAX, needless lifetimes elided, match_like_matches_macro
  applied, io::Error::new(Other) -> io::Error::other, the
  AuthMethod::new self-taking signature renamed
  into_starting_state, the F-4 ToSocketAddrs string re-worded — it
  no longer names the cut resolve_dns(), the unused ready! macro
  and dead AuthenticationMethod::from_u8 removed, the upstream
  commented-out async-std note folded into the impl doc)
- Module docs rewritten: ownership statement (project conventions
  apply unconditionally), extraction provenance (ADR-011) +
  ownership (ADR-013) + maintenance posture (behavioral reference,
  hand-port path)
- Differential suite survives the rename untouched (32 tests pin
  bytes, not upstream text)

Verification: cargo test (38+32 green), cargo clippy --all-targets
-D warnings, cargo fmt --check, cargo check/clippy
--target wasm32-unknown-unknown, cargo test --all-features, cargo
doc --no-deps — all pass.
2026-09-22 12:41:30 +00:00
glm-5.3-flash d6fbd04ef3 vendor: the differential RFC edge-case suite — 32 duplex-only tests
tests/vendor-differential.rs pins the vendored extraction's RFC edge
behavior as explicit bytes, per vendor.md §Maintenance Posture (the
differential-testing contract; ADR-011):

- negotiation: empty/zero method lists -> 0xff, server-order method
  preference, 0x04 greeting refusal before the methods byte, EOF
  mid-handshake typed error, password sub-protocol EmptyUsername/
  EmptyPassword + the 0x01/0xff-then-error reject path
- command reads: unknown command -> 0x07 reply, RSV != 0 tolerance
  (upstream reads it unchecked), ATYP 0x02/0x05 -> 0x08, the
  255-byte domain wire maximum, truncated v4/domain address errors
- reply forms (new_reply regression guard): 10/22-byte shapes,
  constant VER/RSV across codes, zero-addr error replies, caller
  sock_addr verbatim on success
- client session: per-code ReplyError mapping (incl. the 0x06
  double map), domain request encoding, UDP-ASSOCIATE zero-target
  fallback form, skip_auth emits no greeting
- UDP codec: GarbageInReserved, FRAG passthrough (the parser
  returns it; refusing is the caller's job), v6 round-trip,
  payload split
- transfer/wait_on_tcp: pump both directions to completion, EOF on
  clean close, UnexpectedUdpControlGarbage carries the byte

Every non-obvious expectation names the upstream behavior it pins
(file + line in /workspace/fast-socks5, v1.0.0); the reference is
consulted, never linked — no build-time dependency. Duplex pairs
only (no TcpListener/UdpSocket/tokio::net in test code).

Verification: cargo test (37 unit + 32 differential),
cargo test --all-features (identical), cargo clippy --all-targets
-- -D warnings, cargo fmt --check, cargo check --target
wasm32-unknown-unknown — all clean.

Task: tasks/implementation/vendor-differential-tests.md (status ->
completed).
2026-09-21 15:27:20 +00:00