Commit Graph
3 Commits
Author SHA1 Message Date
glm-5.3-flash bc4a9b6024 docs: align remaining alloc-failure references with ADR-010 §2A (prepublish review for v0.3.0)
- README.md ADR index: ADR-010 status notes the §2A amendment
- ADR-009: amendment note records the §2A supersession (in-band path
  shrinks to nothing from registered producers)
- ADR-010 Consequences: dial_failed replaces the stale in-band
  allocate_failed retry-policy bullet
- AllocFailed doc comment + tty-backend.md: describe both failure
  surfaces (direct-path frame / channels-path dial_failed)
2026-09-07 10:04:44 +00:00
glm-5.3-flash af9dbaedea feat: allocation moves into the channels establisher (alkcall 0.6.0 / review 007 R-01) — bump to 0.3.0
- Bump alkcall 0.5.0 -> 0.6.0 (Establishment plan payload, ADR-049
  amendment 2); alktty 0.2.0 -> 0.3.0 (breaking channels-path
  allocation-failure shape)
- make_tty_establisher runs backend.allocate; failure maps to
  EstablishmentError::DialFailed (channel:open_failed reason
  dial_failed); the TtyHandle crosses to the pump handler via a
  private per-open AllocatedHandle one-shot slot (the handle is not
  Sync — it cannot be the ChannelPlan directly)
- make_tty_open_handler gains the Option<ChannelPlan> parameter; a
  plan drives the new drive_session_pre_allocated (pumps only), a
  None plan falls back to drive_session_pre_negotiated (inline
  validate-and-allocate — defense-in-depth for no-establisher
  registrations)
- tty_open_spec's channel:open_failed ErrorDefinition declares
  dial_failed (four reachable reasons)
- Pinned in-band allocate test flipped:
  allocate_failure_fails_open_as_dial_failed (end-to-end, no channel
  survives); establisher unit gate for the plan slot + failure
  mapping
- ADR-010 amended (§2A; §2 kept as historical record), tty-adapter.md,
  session/channels/adapter docs, CHANGELOG, AGENTS.md alkcall pin
  updated

Verification: cargo test (113) + --all-features (137), clippy
-D warnings (native + wasm32, default + all-features), fmt, doc,
wasm32 check — all clean
2026-09-07 09:53:57 +00:00
glm-5.3-flash e2fa32b3c7 feat: channels-path establisher migration (alkcall 0.5.0 / ADR-049) — bump to 0.2.0
Adopt alkcall 0.5.0's channel-open establishment phase (ADR-049 —
review 006 E-01 + N-1) and migrate the channels-path semantic failures
per its §5 sequencing (alktty ADR-010).

- `register_openable` registers `channels/tty/sub` with an establisher
  (`register_openable_with_establisher`): full `NegotiateRequest`
  parse of schema-valid `input`, `carriage == "raw"`, non-empty `cmd`,
  backend lookup, and the ADR-050 ownership check run before the open
  reply; rejections are `channel:open_failed` with `details.reason`
  (`unknown_resource` / `handler_error` / `timeout`) — no phantom
  channel (the SSH contract holds consumer-visibly)
- `backend.allocate` deliberately stays in the pump handler:
  `Establishment` is payloadless so the `TtyHandle` cannot cross the
  establisher→handler boundary, and re-allocating would violate
  ADR-005's kill-on-Drop contract — `allocate_failed` remains the one
  in-band failure class on the channels path (pinned by test)
- `TtySessionError::ChannelsOpen` carries alkcall's typed
  `ChannelOpenError` (`#[from]`) instead of a flattened `String` —
  the N-1 fix at alktty's layer (breaking)
- channels-path semantic failures change shape from
  `NegotiationRejected` in-band frames to `channel:open_failed` call
  errors (breaking); the direct-ALPN path is unchanged
- `tty_open_spec()` gains a `description` (review 006 E-02) and an
  ErrorDefinition for `channel:open_failed` (ADR-016 — disclosed via
  services/schema)
- alkcall = "0.5.0"; version 0.2.0; ADR-010 + ADR-009 amendment +
  tty-adapter.md + CHANGELOG

Verification: cargo test (112 lib + integration), cargo test
--all-features (136), clippy --all-targets -D warnings (host + wasm),
fmt --check, cargo doc --no-deps clean; wasm32-unknown-unknown check
confirms the default crate stays wasm-clean.
2026-09-06 20:25:39 +00:00