Repository navigation
row 155 F25 REHOME (kernel): walk, exhaustion, deadline, probe and breaker rows over the real store, 37 legacy tests; a silent stream head waits under the stream ceiling as 1.5.5 - #905
Merged
Conversation
… probes and breaker over the real store; a silent stream head waits under the stream ceiling, as 1.5.5 37 rows of coverage-map-v2.tsv whose home is a busbar-kernel route/probe/breaker test file, ported from the legacy busbar-llm engine at equal strength over the real breaker store (a private adapter shaped as the root's): fallback-pool probe cells (2xx closes, transport error and 5xx re-open with backoff, 4xx releases without penalty), the max_requests budget (200, 200, 503), exclusions, the permit held through a stream, the pre-first-byte trip, the SSE/buffered refund arms, the at-capacity spills and cascades, the queue that won a permit on an opened breaker, the SWRR spread, the probe classifications (500/503 transient, 401 hard down, the window dilution) and the probe guard on the real breaker. Fix (rows 33-35, red before it): a streamed unit whose member states no attempt cap waited for its head under the walk's failover budget (120 s default) and the connector's open timeout was that budget too. 1.5.5 bounded only a buffered request by the failover deadline; a stream, head included, ran under upstream_request_timeout_secs (v1.5.5 crates/busbar/src/proxy/engine/mod.rs: 1684-1698; the member cap at :1712-1717). far_end.rs attempt_cap and the first-answer wait now take the stream ceiling for such a unit.
MattJackson
added a commit
that referenced
this pull request
Oct 15, 2026
…nel rehome holds a twin of each F25's kernel rehome (#896/#900/#905) ported legacy busbar-llm engine tests into the kernel under the same fn names, so each bare ref in PB-3/4/5/8/27/55/57/73 is declared in two files and settles nothing. Each cites the busbar-llm test it was curated against as path.rs::name, the form PB-5 already uses.
MattJackson
added a commit
that referenced
this pull request
Oct 15, 2026
…ased breaker port; the root rigs on the shared plane handle; law0 back at its row From ModernCI's full logs at 5ad45bf: - busbar-kernel walk_tests / probe_unit_tests (F25 rehome #905/#896): their private breaker adapters classify as the root's BreakerAdapter now does (the transport's fault reading, no number: busbar_kernel_breaker::port::UpstreamStatus { reading }), the four frame_with_upstream calls state their fault and retry_after, the head of answer() carries its class's fault reading as harness::frame does, and route_async answers Pick::Refused as the harness does (E0432/E0425/E0560/E0061/E0004). - busbar root: compose_planes takes the shared Arc<Plugin<Plane>> (#986) in the session door rig and the a2a root cells; DoorReach states catalog in door_steps' own_reach and the a2a root cells (E0308/E0063). - clippy: config_validate_cross_plane uses contains. - kind-isolation:law0 busbar x plane 626 vs 620: the perf test's fixture and the test-linked comment named planes (+6); renamed fixtures/perf_mock_provider.yaml and reworded. - linked-dropped-features: its four transport-http tokio rows are STALE (drained), struck.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
F25 REHOME (status row 155, kernel lane): the legacy
crates/busbar-llmtests whose coverage-map home is a busbar-kernel walk / exhaustion / deadline / probe / breaker test file, ported at equal strength over the REAL breaker store (a private adapter inroute/walk_tests.rsshaped exactly as the root'scrates/busbar/src/root/adapters.rs;probe_unit_tests.rshas its own copy). Literal statuses, Retry-After, cell states, err/window counts, budgets, elapsed times.Fix (rows 33, 34, 35 are RED without it)
A streamed unit whose member states no
attempt_timeout_mswaited for its HEAD under the walk's failover budget (120 s default), and the connector's open timeout was that budget too (far_end.rsattempt_cap, first-answer wait). 1.5.5 bounded only a buffered request by the failover deadline; a stream, head included, ran underupstream_request_timeout_secs(v1.5.5crates/busbar/src/proxy/engine/mod.rs:1684-1698; the member cap at:1712-1717). Both now take the stream ceiling for such a unit. Side effect: a streamed unit's auth-refresh wait sharesattempt_cap()and is the ceiling too.Rows (coverage-map-v2.tsv line numbers; legacy under
crates/busbar-llm/src/engine/), 37 rowsroute/deadline_tests.rs: 33, 35 (engine_tests/send_envelope_tests.rs,engine_tests/stream_deadline_tests.rs)route/exhaustion_tests.rs: 34, 181, 480 (512 seeded worlds; no proptest in kernel dev-deps), 483, 484, 522, 526, 528, 529, 531, 532, 533, 546route/walk_tests.rs: 95, 118, 119, 120, 121, 135 (laneok == 2read as the cell window (2,0): the kernel keeps no lane ok counter), 136, 155, 364, 366, 587route/pick_order_tests.rs: 552, 553busbar-kernel-egress/src/tests/attempt_tests.rs(new, wired fromattempt.rs): 48plane_driver/tests/probe_unit_tests.rs: 189, 221, 222, 234 (moved here fromtests/probe_tests.rs, which cannot reach the probe rig)tests/members/breaker/probe_tests.rs: 225, 226busbar-kernel-breaker/src/tests/mod.rs: 568, 570Not here (behaviour is the root's): 224 (served probe skips only passthrough members; the root binds an own-credential lane with an empty key as a normal auth binding, so it is probed: 1.5.5 never probed a keyless lane,
crates/busbar/src/health.rs:250-254), 602 (crates/busbar/src/root/serve.rs:764always carriesprobe_schedule; 1.5.5 carried it only for an unchanged lane set,crates/busbar/src/main.rs:3596-3606).Files also held by open PRs, hunks kept clear:
plane_driver/far_end.rs(#840 l.1291+, #847/#837/#885/#886 l.1074, drafts #778/#561; this PR edits l.847-866 and l.1154-1185),walk_tests.rs/probe_unit_tests.rs/busbar-kernel-breaker/src/tests/mod.rs(drafts #778, #561).