Three inline findings from the session review of ADR-072/ADR-073:
- fix(auth): Basic door parsing no longer requires UTF-8 — the decoded
password bytes pass through to resolve_from_token byte-faithfully, so
RFC 7617 legacy ISO-8859-1 credentials resolve instead of parsing to
None. ADR-073 already specced raw-bytes passthrough ("decoded and fed
as the AuthToken raw bytes"); the String::from_utf8 round-trip
contradicted it.
- test: pin the with_ws_max_sessions fix against the shape it fixed —
a builder call after with_ws_max_sessions (the pre-router_state()
builder silently reverted the cap).
- test: pin that ownership wiring survives a later HttpAdapter builder
rebuild at the HTTP surface.
Verified: cargo test (451), --all-features (530), clippy --all-targets
-D warnings, fmt --check, fuzz corpus replay 21/21, cargo doc.
review 008 resolution:
- AH-1: door_auth_middleware on extras doors resolves
Authorization: Basic base64(user:pass) — password field as the
token (GitHub-PAT shape, git-lfs / git smart-HTTP clients) — through
the same IdentityProvider::resolve_from_token backend as Bearer; no
new trait method, resolve-fails -> None fallback preserved. Bearer
surfaces stay Bearer-only: the gateway endpoints, /openapi.json,
/healthz, the MCP nest, and the WS upgrade path never resolve Basic.
The 'Bearer-only is the auth mechanism' doc line flips (ADR-073).
- AH-2 ride-along: extras door route tables are static — the merged
router owns the decoy fallback, a second fallback panics at
construction; documented on with_extra_routes and ADR-046 §1.
docs: ADR-073, ADR-046 §1/§4 amendment, ADR-004 presentation table,
review 008 marked resolved; new server::* re-exports; dep:base64
behind the server feature.
Verification: cargo test (449 pass, incl. door Basic resolve-through
and gateway-stays-Bearer-only pins), test --all-features (528),
clippy --all-targets -D warnings, fmt --check, fuzz corpus replay
(21/21), cargo doc.
Ride the alkcall 0.10 gateway ownership seam (review 012, ADR-048
amendment) into the door layer:
- HttpAdapter::with_ownership_provider /
HttpAdapter::with_static_ownership_fallback mirror alkcall's builder
shape; OwnershipProvider re-exported next to GatewayDispatch
- RouterState carries the wiring; GatewayState::dispatch is the single
spine builder for both the gateway routes and the MCP nest
- loud default passes through: a resource_type-gated op on an unwired
spine is refused (FORBIDDEN envelope -> 403/401) on HTTP exactly as
on the native wire
- WS native-session path untouched (consults ownership natively)
fix: with_ws_max_sessions never stored the fresh session-slot
semaphore; later builder calls silently reverted the cap. Builder
state now assembled through one router_state() helper.
infra: fuzz/shared follows to alkcall 0.10 (two alkcall copies in one
build would type-clash); alkhttp 0.8.0 per the dep-wave convention.
Verification: cargo test (444 pass, incl. 3 new call_resource_gated_*
HTTP-surface pins), clippy --all-targets -D warnings, fmt --check,
test --all-features (520), fuzz corpus replay (21/21), doc.
- websocket/upgrade.rs: OpenableAlpn's spec doc advertised the removed
`channel:pub_open_not_implemented` stub (upstream C-08, fixed in
alkcall with the single-publish fixes) — rewritten to the current
contract: Pub-typed marked specs are refused loudly at registration
by `register_openable_with_establisher`; marked pub openables route
through `ChannelCore::register_pub_openable`. Also dropped the
stale 'alkcall 0.6.0' version pin
- adapters/from_openapi.rs: clippy rust-1.99
`cloned_ref_to_slice_refs` — std::slice::from_ref instead of a
single-element to_string slice
Verified: cargo test (458 ok), clippy --all-targets -D warnings,
fmt --check
Two stateful op-sequence harnesses (fuzz/shared):
- ws_session_opseq: the WS byte adapter's generic pumps driven
end-to-end over a mock socket flavor through the new
byte_adapter::fuzzing drive hub (FuzzMockSocket flavor,
FuzzChannels/FuzzWriteSide channel bundle, pump spawn funnels,
FuzzInboundChunkProgress). Asserts lossless inbound forwarding
(exact after the teardown joins), the WS-01 walk's completed-chunk
count vs the declared frame count, the WS-15 close mapping
(1002/1011/verbatim-requested + per-cause reasons), lossless write
framing onto the declared chunk stream (FWD-21's failure-carry
included), lane-failure isolation (the alkcall 0.8.1 class), and
pump termination within the bound.
- retry_throttle: the inlined RetryAfterMiddleware's throttle map
through the new client::fuzzing hub — capacity bound, the
expired-first + Farthest-future eviction policy, the ceiling clamp,
refresh contraction, the Retry-After parse partition, and the
jitter bounds.
16 committed seeds (8 per target) in gen_fuzz_seeds.py, hand-encoded
against the arbitrary 1.4.x derive layout with ceiling-division
variant tags and per-element keep-going bytes, pinned by
seed-decode tests. fuzz/shared gains bytes/futures/tokio/url/
http/httpdate/arbitrary deps.
Verification: main crate 441 default tests, fuzz shared replay 21/21
(default + no-default-features), cargo fuzz build clean (5 targets),
clippy+fmt clean both crates.
FWD-21, found minutes into the first ws_session_opseq campaign: one
feed that drains a complete chunk and then hits an oversize header
(or the pending cap) in the same call returned
Walk::Oversize/PendingOverCap and silently discarded the already
drained pieces — the walk's own consumption-accounting contract
('every parsed byte is either emitted or retained', wave-1
invariant 6) was violated by the crate. Production could not reach
the shape (the mux's per-write invariant caps pending at one chunk +
header), so impact was nil, but WriteChunkWalk is public API and the
wave-1 split_walk equality check was skipped on exactly this path.
- Walk::Oversize(Vec<Vec<u8>>) / Walk::PendingOverCap(Vec<Vec<u8>>)
carry the pre-failure pieces (user-approved semver-relevant API
change)
- run_write_pump emits the carried pieces before fail_write_pump, so
a 1011 close never swallows a legal chunk the mux already counted
as written
- the ws_chunk_framing harness's split-drive failure arm now
reconciles the parsed-before-failure bytes across the whole and
split drives; a regression seed covers the drain-then-fail shape
Verification: main crate 441 default / 520 all-features tests, fuzz
shared replay 21/21 (default + no-default-features), cargo fuzz build
clean, clippy+fmt clean both crates.
Adopt the alk*/fuzzing posture (alkcall 0.8.1 precedent) — the dep
bump this ride-along builds on was 4005517:
- docs/plans/fuzzing.md: the four-target plan (sse_parser,
ws_chunk_framing this session; spec_import + the stateful wave
later), the infrastructure blueprint, and the corpus/campaign
policies (§3's table tracks progress)
- fuzz/ workspace (the sibling layout): nightly-pinned subtree,
stable fuzz/shared invariant library (corpus replay = the standing
fuzz gate), detached runner (never a foreground child of an agent
session), JSON dict, deterministic seed generator; 80 committed seeds
- two wave-1 targets: sse_parser (SseParser::feed + the FWD-17
envelope mapping — the parser every subscription forwards through)
and ws_chunk_framing (WriteChunkWalk, the write pump's chunk walk)
- crate-side surfacing: SseParser/SseEvent/sse_event_envelope are pub
behind the adapters::forward::fuzzing re-export hub; the write
pump's inline chunk loop extracted as WriteChunkWalk (pump byte
behavior unchanged — 441 tests green); retained_bytes() diagnostic
(the alktunnels/alksocks posture)
- root Cargo.toml gains the explicit [workspace] table with
fuzz/ excluded (auto-discovery would pull the nightly-consumed
dev-deps into the stable build — the alkcall trap); fuzz/ joins the
publish exclude list
- AGENTS.md verification checklist gains the corpus-replay line
FWD-20 fix, found by driving the new sse_parser harness: feed()
rescanned the whole reassembly buffer per feed — a dribble-fed
subscription (one unterminated line over many small chunks) cost
quadratic CPU (1 MiB dribble ≈ minutes per connection, unbounded
total). Line assembly now walks only the arriving chunk: the buffer
holds just the current line, the scan never revisits drained bytes.
1 MiB dribble: ~5 min (quadratic) → 21 ms (linear). Regression test
sse_parser_dribble_feeds_do_not_rescan_the_buffer.
Verification: cargo test 441 passed; cargo test --all-features ok;
clippy --all-targets -- -D warnings (main + fuzz/shared) clean;
fmt --check (main, fuzz, fuzz/shared) clean; cargo fuzz build clean;
corpus replay (cargo test --manifest-path fuzz/shared/Cargo.toml) 80
seeds green.
Refs: alkcall docs/research/fuzzing.md (§7.9 posture), alksocks
08a3694 (the sibling wave-1 shape), docs/plans/fuzzing.md §2
- the 1.85 claim was already false at the dependency level: the resolved
lockfile pulls icu_* 2.x (MSRV 1.86 via jsonschema -> idna) and, since
alkcall 0.7.1, alkcall itself declares 1.88 — 1.85 toolchains cannot
build the tree regardless of what Cargo.toml declared
- raising the floor to 1.88 aligns with the ecosystem MSRV decision
(noq QUIC path, per the alknet MSRV audit; recorded for alktls in
OQ-TLS-08) and breaks no downstream that could build the crate before
- alkcall 0.7.1 rides in via the lockfile (version = "0.7" already
covered it); no code change forced — 0.7.1 is its own MSRV raise plus
the cf-006/cf-007 work consumed in the 0.7.0 wave
- one 1.88 clippy lint fixed: uninlined_format_args in a forward.rs test
Verification: 1.85 check fails as documented (icu_* 1.88 floors); 1.88
toolchain check/test/clippy --all-targets -D warnings clean; stable 1.94
test 456 (default) / 589 (all-features), clippy -D warnings (default +
all-features), fmt --check, doc --no-deps
Dependency wave + one wire-surface behavior fix:
- bump alkcall 0.6 -> 0.7 (gateway feature unchanged). No code change
forced: this crate constructs no ServingConfig (the one 0.7.0
struct-literal break — from_wss keeps the pure-consumer
from_connection; opt-in serving remains the review-006 Part C
deferred decision), and the CF-006 per-call opener-identity overlay
is behavior-only upstream (hub-forwarded WS-ferry opens now resolve
the end client's identity, the intended effect)
- gateway error table maps CONNECTION_CLOSED -> 503 Service
Unavailable: alkcall 0.7.0's CF-007 amendment makes it the eighth
protocol code (ADR-016's list is formal at eight); it previously
fell to the catch-all 500, misreporting a retryable,
provably-undelivered call as a permanent server failure. retryable:
true, so the retry_after details path applies on 503. Tests pin the
mapping and the Retry-After passthrough
- ADR-023 amended: the protocol-code list is eight (ALREADY_EXISTS +
CONNECTION_CLOSED rows added to the §3 table, normative rule and
call.error code list updated); http-adapters.md six-code references
de-staled
Verification: cargo test (456 passed), cargo test --all-features
(589 passed), clippy -D warnings (default + all-features), fmt
--check, cargo doc --no-deps (0 warnings)
Mechanical pass for the alkcall 0.6.0 bump (review 007's
establishment follow-ups sweep: R-01 Establishment plan payload,
R-02 OpenHandler lifetime doc note, R-03 pump_bidi extraction):
- bump alkcall 0.5 -> 0.6 (gateway feature unchanged)
- the OpenHandler signature gained a plan parameter —
Fn(Value, Option<ChannelPlan>, Connection, AuthContext) ->
JoinHandle<()> — so the test's echo_open_handler closure gains
`_plan` (the only closure-construction site in this crate)
- no ferry change: this crate constructs no Establishment and passes
the Option<OpenEstablisher> through unchanged (OpenableAlpn fields
and register_openable_with_establisher threading are unchanged in
0.6.0); the establisher's Establishment.plan now reaches the pump
handler's second parameter process-locally (ADR-049 amendment 2)
- docs updated: websocket.md (plan threading + pump_bidi pointer),
OpenableAlpn field docs, with_ws_openable_alpns doc comment
- CHANGELOG [Unreleased] consumption-wave entry
Verification: cargo test (454 passed), cargo test --all-features
(587 passed), clippy -D warnings (default + all-features),
fmt --check, cargo doc --no-deps clean
cargo doc --no-deps was not actually clean before review 007's
remediation — it emitted 5 warnings that went uncounted (grep for
"^error|^warning" in the gate runs matched only on error/warning at
line start, and the review baselines recorded "clean" without a
counting check; the warnings were first noticed while re-verifying
the review-007 gate set). All pre-existing, none introduced by
review 007:
- src/adapters/mod.rs: the bare [`OpenAPISpec`] intra-doc link did
not resolve (the re-export sits in the same module, but rustdoc
resolves module-level links against the module's own scope) —
spelled out as `openapi_spec::OpenAPISpec`
- src/websocket/mod.rs: [`upgrade`] did not resolve from the
`websocket` module doc (the submodule is not in scope for the
parent's doc links) — plain `upgrade`
- src/websocket/upgrade.rs:215: [`CallConnection`] had no import in
scope — spelled out via the full alkcall path
- src/websocket/upgrade.rs:291: [`install_channel_zero`] linked a
private item from public docs — plain `install_channel_zero` hook
- src/server/adapter.rs:274: [`crate::websocket::OpenHandler`] is
not re-exported from `websocket` — spelled out via the alkcall
path
Verification: cargo doc --no-deps 0 warnings (was 5); cargo test 454
passed / 0 failed; cargo test --all-features 587 passed / 0 failed;
clippy (both configs) clean; fmt clean.
Review: docs/reviews/007-ws-data-channel-surface-review.md (the
baseline's "cargo doc → clean" claim was wrong on this dimension)
Review 007 Unit 2 (the two "implement" decisions taken during
remediation, plus the coverage gap):
- WS-29: the `op/register` override surface review-006 UP-02 and
ADR-048 recorded as landed is now implemented. The hook threads an
`op_register_acl` `AccessControl` into `op_register_spec` (the
permissive `AccessControl::default()` remains the default
everywhere); the built-in surface sets it via
`HttpAdapter::with_ws_op_register_acl`, bare-registry/custom routes
via the `OpRegisterAcl` request extension (mirroring
`ChannelsPolicy`/`WsTimeouts`/`OpenableAlpns`). A peer whose
identity does not satisfy the ACL gets `FORBIDDEN` on the announce.
Gates: builder path (`FORBIDDEN` scope-less / announce-ok scoped)
+ extension path.
- WS-30: the bare-registry `SessionState` is built by `FromRef` per
request, so its default-cap semaphore bounds nothing across
requests (corrects review-002 WS-17's "bounded at 64 sessions"
claim). New `SessionSlots` request extension carries the shared
semaphore for routes that need an effective cap; the upgrade
handler prefers it over the state value. Doc comments corrected
(`SessionState`, `WsTimeouts`, `ws_upgrade_handler`). Gate:
cap-1 route → 503 over cap → slot freed on session end.
- WS-32: the built-in openables threading
(`with_ws_openable_alpns` → `RouterState` → `SessionState` → hook)
gets its first gate — every Unit-3 gate rode the `OpenableAlpns`
extension fallback. `builder_path_openables_serve_the_data_channel_
surface` discovers the openable via `services/list`, opens the
channel, and round-trips bytes through the builder-built router.
Verification: cargo test 454 passed / 0 failed; cargo test
--all-features 587 passed / 0 failed (+5 gates); clippy (both
configs) clean; fmt clean.
Review: docs/reviews/007-ws-data-channel-surface-review.md
The WS-26 `ConnectionGuard` was bound inside its `if let` block, so it
dropped microseconds after insertion instead of living for the
channel-0 dispatcher task — `live_connections()` /
`live_connection_count()` were permanently empty for every session
(review 007 WS-28 [major], reproduced empirically; the handle
retention ADR-048 + review-006 record was aspirational at that commit).
Remediation (review 007 Unit 1):
- bind the guard as an `Option<ConnectionGuard>` in the channel-0
task's frame, mirroring `SessionGuard`'s shape in
`run_channels_session`; the block comment now describes the real
scope
- gate `live_connections_visible_mid_session_and_drain_after_teardown`
— handle visible mid-session (after a completed call proves the
dispatcher is up), drained after teardown; verified to fail against
the pre-fix tree and pass with the fix
Verification: cargo test 454 passed / 0 failed; cargo test
--all-features 582 passed / 0 failed; clippy (both configs) clean; fmt
clean.
Review: docs/reviews/007-ws-data-channel-surface-review.md (WS-28)
The WS path wires the alkcall 0.3 per-session mechanisms — the OQ-05
deferred half (review-003 WS-20/21/22/25/26, the decisions WS-24/WS-25
resolved upstream):
- install_channel_zero reworked per the ADR-047 §4 amendment #2 shape:
fork the deployment's base registry, register the generic channel
ops (channel/close, channel/control, channel/resources/subscribe —
WS-21), the deployment's openable ALPNs (ChannelCore::register_openable,
WS-22), the bootstrap discovery set closed over the fork
(install_bootstrap_discovery, F-06), and op/register (WS-25; the
collision set is the fork per review-005 G-03), then dispatch over
the fork. The session's ChannelsPolicy rides the hook (one policy
instance across open wrappers and the demux teardown path).
- OpenableAlpn { spec, open_handler } + HttpAdapter::with_ws_openable_alpns,
threaded RouterState -> SessionState -> hook, with the OpenableAlpns
request-extension fallback (mirroring ChannelsPolicy/WsTimeouts).
- WsSessions retains the channel-0 Arc<CallConnection> (WS-26) with a
self-removing guard (ConnectionGuard); live_connections() is the
deployment-visible surface.
- UP-01: ALREADY_EXISTS maps to 409 Conflict in the gateway error map.
- from_wss import excludes the protocol-session ops (bootstrap set +
channel lifecycle ops): the fork serves them per session, and proxying
session-scoped machinery (e.g. channel/close across sessions) would be
nonsense. Discovery runs first, the filter is the listing minus those
names.
- adapter_install_channel_zero cfg matches its caller (WS-27); it
inherits the reworked hook (session ops now served in the from_wss
test-server producer too).
Gates (Unit 3, tests/ws_upgrade_session.rs; the WS-23 e2e set):
open -> channel_id -> discoverable in services/list -> chunks both
ways -> handler sees bytes; channel/close resolves + ledger decrement;
cap denial (channel:-prefixed); mid-open disconnect teardown; TooLarge
demux resync through the WS path (16 MiB + 1 skip consumed);
op/register announce + overlay-collision + serving-registry-collision
ALREADY_EXISTS through the WS path.
call_and_await now filters by request id and tolerates data-channel
chunks (a prior Sub's trailing call.completed may interleave).
Verification: cargo test 454 (default) / 582 (all-features), clippy
both sides -D warnings clean, fmt clean, doc clean.
Bump alkcall 0.2 -> 0.3 (published 2026-09-04; the remediation of
alkcall reviews 004-005, the upstream resolution of alkhttp review 003
WS-24/WS-25). Two source-break classes, both mechanical:
- take_call_connection now returns Option<Arc<CallConnection>>
(was bare value) — drop the double-wrap in WssSession::connect.
- register/register_on/register_openable take &OperationRegistry
(was &mut) — drop 51 stale `mut` bindings (clippy --fix), which is
the interior-mutability change that makes the per-session fork
surface possible.
No behavior change in this crate yet; the 0.3 mechanisms (fork,
serve_single_stream, op/register, install_bootstrap_discovery,
collision policy) are the composition surface for alkhttp review 003
Unit 2 (WS-20..22, WS-26 wiring) — not wired here yet.
Verification: cargo test 453 passed (default) / 575 passed
(all-features), clippy both sides -D warnings clean, fmt clean,
cargo doc --no-deps clean.
Split the feature graph so consumers pulling only the import adapters
(from_openapi / from_jsonschema / from_mcp) no longer compile the axum
/ hyper server stack, and server-only deployments no longer compile
reqwest. One crate, one import path — sides cut by features, not by a
crate split.
Feature graph:
- server (default): axum host, gateway, WS upgrade, to_openapi, to_mcp
- client (default): client host, forward, from_jsonschema, from_openapi
- openapi: shared OpenAPISpec model (implied by both sides)
- mcp: from_mcp needs client, to_mcp needs server
- wss: tungstenite transport (from_wss); tungstenite half of the
shared WS↔byte-stream adapter
- h2/http1: hyper protocol features; imply server
Wire-contract neutral: gateway endpoints, ALPNs, and all public API
shapes unchanged; defaults keep both sides on.
Supporting changes:
- forward.rs drops its axum::body::Bytes type leak (bytes crate types)
- bounded_join + error-echo caps move to input_validation (usable by
both sides; openapi_spec no longer imports from forward)
- byte_adapter: axum flavor compiles under server, tungstenite under
wss; the generic pumps stay shared (WS-11)
- input_validation / openapi_spec import-only internals gated to the
side that consumes them
- http-body-util moves to dev-dependencies (was test-only)
- integration-test required-features updated for the new sides
- from_wss unit tests (axum producer harness) gated to server
Verified: cargo test (defaults, 453) and --all-features (575) pass;
lean side builds (client / server / client,mcp / client,wss /
server,wss / openapi-only) build clean with zero warnings;
clippy -D warnings clean across all feature combinations; fmt clean.
Bump the alkcall dependency to 0.2 (with the gateway feature) and
converge on the promoted shared pieces:
- The local dispatch spine (gateway/dispatch.rs, 721 lines) is deleted;
GatewayDispatch, schema_disclosure_denial, and DEFAULT_DEADLINE are
re-exported from alkcall::gateway (alkcall ADR-048). The 30 s default
deadline preserves the previous behavior exactly.
- gateway/schema_cache.rs (PublishSchemaCache) is deleted: alkcall CF-003
compiles publish_schema at registration time and exposes
OperationRegistry::publish_validator; the /publish chunk stream
resolves against it. Un-compilable schemas are now rejected at
registration, so the two end-to-end fail-closed tests were reworked
into a registration-rejection test (a stronger guarantee).
- schema_disclosure_denial consumers (to_mcp, routes) use alkcall's
promoted implementation; the alkhttp-local copy is gone (ADR-071
updated: the guard stays as defense-in-depth, the implementation no
longer forks).
- CF-001: from_wss drop monitor and the WS overlay tests use
CallError::connection_closed; the review-001-ws-eof-signal race tests
now assert retryable CONNECTION_CLOSED on both resolution paths (the
tolerated non-retryable INTERNAL write-failure outcome is gone).
- Added CHANGELOG.md (Keep a Changelog), Unreleased section records the
bump and convergence.
Verification: cargo test default 453 ok, wss 470 ok, mcp 526 ok,
all-features 575 ok; clippy -D warnings clean (default + all-features,
all-targets); fmt clean; cargo doc warning-free.
Net: -1093 lines.
Decision: advertise == enforce. The key allowlist (OAI-02) stays as the
first gate with its established unknown-key message; a compiled leaf
validator now runs second, so required/type/enum/pattern/bounds
violations surface as INVALID_INPUT 422 naming the keyword — not as
upstream round-trips.
- new src/adapters/input_validation.rs: CompiledInputSchema compiles an
op's input_schema once at import with the jsonschema crate (same
2020-12 dialect publish_schema uses) and validates peer input at call
time; the compile-time copy is hardened closed-by-default
(additionalProperties: false injected when absent) so the validator
reproduces the allowlist's unknown-key semantics; explicit
additionalProperties:true catch-all and schema values are preserved;
the original spec value is never mutated
- from_openapi/from_jsonschema import(): compile per registration,
capture the validator in the handler closure (re-import recompiles —
the closure capture is the invalidation story); a non-compilable
input schema fails import loudly (AdapterError::SchemaParse naming
the operation), matching the publish_schema fail-closed precedent
- from_openapi generated input schemas now carry
additionalProperties:false explicitly, so the /schema advert states
the enforced rule and external schema-driven validators reach the
same verdicts
- forward/forward_stream/build_request gain an
Option<&CompiledInputSchema> parameter; enforcement runs after the
allowlist
- round-trip test (review 002 Test-gap 10): the /schema-exported
input_schema is compiled with the same validator and driven against
build_request over a 10-input violation matrix — accept-sets exactly
equal in both directions; the chain-test that lets advertise/enforce
drift surface as a CI failure
- ADR-066: new decision section (advertise==enforce) with the trust-
boundary reasoning and the rejected option (b) rationale
- module + enforce_input_schema docs updated to the two-gate shape
cargo test --all-features 596 pass; clippy --all-features/-D warnings,
fmt, doc gates clean.
docs(tasks): mark review-002-fu-oai18-decision done
- forward_stream build-error arm (forward.rs): wire-level test asserts
one INVALID_INPUT envelope then stream end with zero upstream
contact (a panicking responder counts as the contact guard), plus
the from_jsonschema integration mirror (undeclared key + non-scalar
placeholder, each naming its rejection source)
- PEM read-failure arms (http_client.rs): nonexistent CA path →
CaBundleRead (sync new), nonexistent client-cert path → ClientCertRead
(async reload, prior generation retained)
- Over-cap poll_write rejection leg (byte_adapter.rs): cap+1 write →
InvalidData naming the cap; stream stays usable for an at-cap write
afterwards
- SSE parser edges: CRLF split across feed chunks frames one line;
invalid-UTF8 data lines drop without killing the frame stream
- from_value structural rejects: non-object doc, missing `info`,
missing `paths`, non-object `paths` each name the member
- Connection-failure arms: accept-path ConnectionClosed →
HandlerError::ConnectionClosed via stream_error_to_handler; read-pump
demux-gone break ends the pump when the byte-stream side is dropped
- Delete the caller-less `impl Default for WsTimeouts` (the extension
is constructed explicitly)
cargo llvm-cov --all-features: all named arms covered; TOTAL regions
94.18% (was 93.86%), lines 96.04% (was 95.77%); http_client.rs
86.56% lines (was 81.72%).
docs(tasks): mark review-002-fu-stream-error-coverage done
- dispatch.rs: qualify [invoke]/[invoke_streaming] as
[GatewayDispatch::invoke]/[::invoke_streaming] (module-level docs
need the type-qualified path)
- byte_adapter.rs: drop redundant explicit link targets for
DEFAULT_WS_WRITE_TIMEOUT / DEFAULT_WS_IDLE_TIMEOUT (already
re-exported at crate::websocket)
- from_wss.rs, byte_adapter.rs: backtick WssSession::drop and the
private WsFraming (all-features-only link errors, same class)
- Ws pump/timeout module docs spot-checked against WS-13/18/17 impls:
no drift found
RUSTDOCFLAGS="-D warnings" cargo doc --no-deps now exits clean both
default-features and --all-features.
docs(tasks): mark review-002-fu-doc-warnings done
build_request inserts header params first, then default_headers, then
credential headers — HeaderMap::insert replaces, so a declared
in: header parameter whose name matched a default or credential header
silently never delivered the peer's value upstream.
check_header_param_collisions runs at import (the assembly's
HttpServiceConfig is visible there): Authorization on an authed
namespace is rejected outright, a default_headers name match (compared
case-insensitively) fails import naming both keys, and an ApiKey
header_name match fails naming the credential header. Per the task's
decision note: the import-time surface does see the adapter's config,
so the loud point stays at import rather than first-call warn-once.
Tests: Authorization+Bearer, X-Tenant/x-tenant case-insensitive
default_headers, x-api-key/X-API-Key, and the no-collision clean path.
Internal/ACL-restricted op specs were readable through
POST /call {"operation":"services/schema","input":{"name":...}}
(the MCP call/batch tools identically): the outer-name pre-checks pass
(services/schema is External) and alkcall's services_schema_handler
projects any registered spec with no visibility/ACL check of its own
(alkcall CF-004 is the complete fix there).
- GatewayDispatch.invoke/invoke_streaming now apply the GET /schema
route's is-internal + access-control checks to the meta-op's inner
name input before dispatch (404 Internal / FORBIDDEN ACL), one
interception point covering /call, /batch, /subscribe and the MCP
call/batch tools; /publish cannot reach the Query-typed meta-op
- the visibility+ACL check is one shared fn (schema_disclosure_denial)
used by the HTTP /schema route, the dispatch guard, and the MCP
schema tool, so transports cannot drift
- when CF-004 lands, this guard remains as defense-in-depth (ADR-071)
Tests: dispatch-spine guard unit tests; /call 404 + 401/403 matrix,
/batch NOT_FOUND entry, /subscribe error event; MCP call/batch tools
via services/schema with an Internal inner name (mcp feature).
Verify: cargo test (405), --all-features (523), clippy default and
--all-features --all-targets -D warnings, fmt --check — all pass.
Import errors echoed unbounded spec-derived strings: a 100k-path
servers-override list produced a multi-megabyte SchemaParse message.
forward::bounded_join caps list echoes at 8 items / 128 chars per item
with a ', … (+N more)' suffix, and is applied at the servers/callbacks
/security location lists, the unbound-placeholder and unbound-remnant
lists in build_registration/forward, and the call-time declared-keys
echo. resolve_ref caps interpolated $ref strings at 128 chars.
Tests: 100k-path servers fixture asserts message < 4 KiB (linear,
completes fast); bounded_join unit test pins count+width truncation
with unchanged small-case output.
GW-16: empty body / malformed first line / missing header fields /
per-line cap / batch over-cap rejections now route through
call_error_to_http_response_with_identity, mapping INVALID_INPUT to
422 — same status as mid-stream chunk errors. One error class, one
status.
GW-17: invoke_sink wraps the registry sink invoke in the same 30 s
tokio::time::timeout the Once-op invoke uses; a hung sink handler
surfaces as a TIMEOUT (504, retryable) error envelope instead of
holding the HTTP connection forever. The sink wrapper bounds the
whole dispatch (chunk pacing included), matching http-server.md's
deadline contract.
Docs: http-server.md error table documents the 422 triggers and the
sink deadline; http-adapters.md batch cap status corrected.
to_openapi: gateway spec version 1.3.0 -> 1.4.0 (ADR-045 minor):
/publish framing faults and /batch cap reject documented at 422 (the
400 slots moved with the runtime); /publish 400 slot removed; 504
description covers the sink dispatch.
Verification: scripts/verify.sh OK (397 tests); cargo test
--all-features OK (513 tests); clippy --all-features --all-targets -D
warnings OK; cargo fmt --check OK.
The OAI-06 loudness matrix had holes: callbacks, security requirement
blocks, top-level oneOf requestBodies, and schema-level
discriminator/xml keywords all vanished silently at import.
- callbacks + security: OpenAPISpec::validate_import_loud_features
(new, run from FromOpenAPI::import) rejects with locations and
remediation. Scoped to the service-import path, not the shared
from_value parse — the published gateway doc round-trips through
from_value inside to_openapi and legitimately declares security
markers for its external clients.
- top-level oneOf requestBody (no content map): refused in
parse_operation as an unresolvable/content-less body (OAI-15 arm,
message names OAI-14 oneOf).
- discriminator/xml inside consumed schemas: per-operation
tracing::warn listing the ignored keys (JSON-forwarding-only stance).
Tests: op-level callbacks, doc-level + op-level security, oneOf
requestBody each fail import naming the feature and location.
- new tests/retry_policy_wire.rs: POST+500 = 1 upstream hit (method
gate), GET 500/503/200 = 3 hits (retry-to-success), budget exhaustion
stops retries under a 50-attempt cap (wall-time + hit-count bounds)
- forward_stream terminal arms on the wire: oversized SSE line ->
one INTERNAL terminal envelope; mid-stream transport abort (staged
via a notify gate so the abort is genuinely mid-stream) -> terminal
envelope after the delivered frame; pending event EOF-flush; dead
port through both forward() and forward_stream()
- Once-path decode arms: malformed application/json 200 -> INTERNAL
decode envelope; application/octet-stream 200 -> byte-array envelope
- COV-12 credential arms: ApiKey and Basic malformed values fail
loudly without echoing secret material; declared header-param
invalid name/value rejections
- new tests/client_config_reload.rs: config() reflects a reloaded
config (FWD-12 atomicity half)
$ref siblings are ignored under OpenAPI 3.0 semantics but would apply
under 3.1 — a 3.1-authored constraint beside a $ref previously
vanished silently, overstating /schema. warn_ref_siblings now fires at
each $ref consumption point (operation parameters, requestBody,
path-item parameters) naming the location and dropped keys, and the
module doc records the version stance: no openapi 3.1 gate, 3.0
reading with the warn as the visibility mechanism.
A requestBody $ref that could not resolve (or resolved to a shape
without a content map) previously turned into a content-less Operation
registered silently — every call would INVALID_INPUT on the gateway
body key. parse_operation now treats an unresolvable request-body ref
and a resolved-but-content-less body the same unfaithful-modeling way
as an unresolvable parameter ref: the operation fails import via the
OAI-04 'unresolvable $ref' arm, whose message now names the
requestBody case (OAI-15) alongside the parameter one.
Tests: self-ref requestBody, missing-component requestBody ref, and
content-less component requestBody each fail import loudly.
- same-host 302 followed with the credential header arriving at the
followed hop (FWD-03's scrub is cross-host only)
- cross-host 302 surfaced as HTTP_302 with the attacker endpoint
receiving zero requests (load-bearing FWD-03 property, pinned)
- redirect loop trips the hop cap -> loud INTERNAL transport error
Import-time path-template checks now run on both adapters:
- from_jsonschema: validate_path_template moves to forward.rs (same
checks, same messages, shared implementation)
- from_openapi: build_registration calls it before building each op, so
a template like /x{open fails import with 'unterminated placeholder'
instead of a per-call INTERNAL on first invoke
- new test: unterminated_path_template_fails_import_not_first_call
assert_eq! in library code was a panic-family residue in the import
collision check (review 002 JS-03). The three vectors are built in
lockstep by import(); zipping them preserves the pairwise walk without
the panic path.
A logical request's Retry-After waits are now bounded by
max_total_retry_duration, and a retry storm can no longer re-arm a
full ceiling per attempt:
- BudgetClock anchored per logical request by RetryGateMiddleware
(into the request Extensions, shared across retry attempts) and
read by the inner RetryAfterMiddleware every attempt; the monotonic
anchor projects the hard stop through wall-clock steps.
- maybe_sleep_for truncates the sleep to the remaining budget;
a spent budget skips the sleep entirely.
- record() keeps the EARLIEST deadline per URL (retry storms cannot
extend the first-seen deadline); a refresh that cannot make it
under the budget hard stop drops the entry so the next attempt
starts immediately instead of parking.
- Middleware without a budget anchor (budget = 0) keeps the prior
semantics; the shared client now wires
HttpClientConfig.max_total_retry_duration into the Retry-After
middleware.
- Wire tests (tests/retry_after_budget.rs): always-429 responder with
a 300 s Retry-After is bounded by budget + one attempt; separate
logical requests still honor the recorded throttle window.
- FWD-12 atomic reload pairing and the FWD-15 stream-client split
untouched (stack built in build_client_with_pems for both).
- assemble_request_url now verifies after Url::set_path that the decoded
URL path segments are byte-identical to base_dir + decoded rendered
segments; a mismatch (any future normalizer rewrite in the url crate)
fails loudly with INTERNAL instead of silently re-routing an
authenticated request
- property-style corpus test over dot/percent/binary values: accepted
values must survive byte-identical as one literal segment with no lone
dot segments; failures may only be INVALID_INPUT
Verification: scripts/verify.sh (383 passed) and --all-features (499
passed), clippy -D warnings, fmt --check all pass.
- value_to_path_segment now rejects scalar values whose decoded form is
exactly '.', '..', '%2e', '%2E', '%2e%2e', or '%2E%2E' (case-insensitive)
with INVALID_INPUT: url 2.5.8 Url::set_path silently normalizes lone
dot segments away, so such values would route to a different upstream
endpoint than the template describes, with namespace credentials attached
- rejection is exact-match on the full decoded segment: dotted values
like v1.2.3, .hidden-file, ..hidden, ... still render
- error names the failure mode but never echoes the raw value
- empirically pins the set_path normalization behavior in a test
(tenants/../resources -> /resources, /files/.. -> /, %2E%2E -> normalized)
Verification: scripts/verify.sh (382 passed) and --all-features (498
passed), clippy -D warnings, fmt --check all pass.
The FWD-08 remediation made every malformed credential fail loudly, but
an authed operation whose registry capability was entirely absent fell
through the build_request match: the request was sent with no credential
and no diagnostic, producing corrupted upstream 401s at call time.
- build_request now returns an INTERNAL error naming the missing
capability keys (api_key:{ns} / http_token:{ns}) when an auth scheme
is declared and Capabilities::get is empty; the request is not sent
- auth_scheme: None behavior unchanged (unauthenticated ops stay
unauthenticated); error message carries key names only, no secret
- module doc: loud-missing matrix now covers malformed name/value AND
absent capability
- tests: unit loud-error across all three schemes, unchanged-arm pin,
wire test asserting the upstream receives zero requests (mirrors the
FWD-08 test family)
- from_openapi/from_jsonschema no_env_vars tests updated: they pinned
the old silent fall-through; still assert no env material echoes
Verification: cargo test (380 passed), cargo test --all-features
(496 passed), clippy --all-targets -D warnings (default + all-features),
cargo fmt --check — all via scripts/verify.sh
glm-5.3-flash