Skip to content

Align artifact transfer deadlines and stall budgets #4441

Open

Description

@thymikee

Current execution status — 2026-11-13

Open; separate reviewed remote and source-route components are unpublished. Ingress/proxy/ticket/disconnect completion, component integration, isolated proof and pinned exact validation remain. Local focused observations are not whole-issue readiness.

Latest detailed Current state.

Parent: #4368. Audit: DSRV-5, CMD-5, CONTR-12. Authorized by the user's acceptance of the next-ten recommendation. Source baseline: 5781ba4aa63462d7fa55b5987e5841a01c691dfc.

Problem and owner

The daemon and proxy HTTP constructors leave whole-request timing to Node defaults, install-source uses a fixed120s transfer signal, and remote upload attempts use a fixed300s timer. These limits can reject progressing transfers independently of the admitted operation. Review the current transfer/cancellation/installation owners before treating any old audit fallback claim as still present.

Owners: packages/daemon/src/server/http/http-server.ts; packages/proxy/src/daemon-proxy.ts; packages/remote/src/upload-stream.ts and upload-client.ts; packages/provision-kit/src/install-source.ts and install-source-download.ts.

Required contract and prevention design

Define the aggregate transfer/installation budget, header/connect budget and inactivity budget explicitly at the existing owners. Progress must not silently reset an aggregate deadline. A progressing transfer inside its admitted budget must not fail because of a contradictory120s/300s sublimit. A stalled, oversized, cancelled or expired transfer must terminate promptly with typed evidence and bounded cleanup. Preserve HTTP framing, ticket expiry, credential boundaries, resume offsets and the existing aggregate installation/result margins.

Use the existing caller/request deadline where available. Before implementation, record how ingress and pre-RPC uploads without that deadline obtain a finite budget and how that budget relates to upload-ticket expiry. This policy decision is required; setting all timeouts to zero, broadening retries or adding an unbounded fallback is not approved. Reuse existing deadline/progress owners rather than a new transfer framework.

Completion evidence

Real loopback client→proxy→daemon transfer tests distinguish slow progress, stalls, abort and aggregate expiry. Exercise install-source response-body consumption and request framing. Retain byte caps and prove no resumed/fallback work starts after cancellation. Use controlled clocks or transport seams rather than minutes-long sleeps. Demonstrate regression against the named base or a meaningful policy mutant. Independent review, full repository format, exact-head affected validation and CI are required. Preserve #4050. No native device or real cloud transfer is required.

Sequencing

The coordinator owns the budget decision and current overlap check before an implementation claim. This issue is not yet an AFK implementation contract. No merge, release, upload or tags are authorized.

Activity

  1. apex-sentinel-staging commented on Nov 11, 2026

    @apex-sentinel-staging

    Apex Sentinel · Issue review

    Suggested label: bug

    The report describes existing fixed 120s/300s transfer timers and Node defaults causing progressing transfers to fail, which is broken behavior. It names affected files, required contract, tests, and sequencing, though it notes the issue is not yet an AFK implementation contract and needs a budget policy decision first.

    Required sections

    • Steps to reproduce (missing)
    • Expected and actual behavior (missing)
    • Version or environment (missing)
  2. thymikee commented on Nov 11, 2026

    @thymikee
    MemberAuthor

    Decision: policy investigation ownership:

    An isolated source-only Luna design task now traces the existing transfer, upload-ticket and request deadline owners. It must propose a finite aggregate/connect-header/inactivity policy, especially pre-RPC/ingress budgets without a caller deadline. This is investigation only. The coordinator records the selected policy before implementation; no source changes or public behavior decision is delegated. Tests are paused outside the single #4425 base comparison.

  3. added
    enhancementNew feature or request
    and removed on Nov 12, 2026
  4. thymikee commented on Nov 12, 2026

    @thymikee
    MemberAuthor

    Decision: finite transfer budgets and the shared source-install envelope

    Adopt this policy for #4441 under the user's active completion authorization. Baseline is eec35e9. The earlier 480-second reservation is rejected. It would leave only 90 seconds for a download. The numbers below are explicit new policy choices where they differ from current behavior.

    1. Keep the existing RPC envelopes. A normal install has 480 seconds. An install-from-source has 600 seconds. Keep the existing 30-second result margin. Do not reserve the entire normal-install envelope before a source download. At source-install admission, enforce the earliest registered execution/result deadline through the existing request cancellation owner. The default source work budget is therefore 570 seconds from admission, including readiness, download, extraction and installation. Each phase consumes that same remaining budget. Keep the existing per-platform native installation caps; Android and the inspected Apple native installer each cap installation at 120 seconds. A native operation also observes the shared request signal, so its effective cap is the earlier native cap or remaining request deadline. Do not reset the operation deadline when download or installation starts. A request without transport metadata still gets a finite source-install execution budget from its descriptor. A standalone source-materialization call gets a 600-second aggregate at entry, combined with caller cancellation. Do not silently turn that API into an unbounded download.

    2. Give remote artifact preparation one 600-second aggregate, beginning before packaging or hashing and ending after the last upload/finalize needed for that preparation. The standalone uploadArtifact API gets the same finite aggregate at entry. An enclosing preparation/caller/lease deadline always wins. All attempts, redirects, offset resumes, legacy fallback, response consumption and finalize share the original aggregate. No retry starts after cancellation or expiry. Wire the signal into packaging children, hashing pipelines, upload pipelines and response readers. Abort and join owned work before cleanup; do not race and abandon it. This preparation budget precedes the RPC envelope, matching the existing lifecycle: a default remote normal install can spend at most 600 seconds preparing plus its existing 480-second RPC envelope; source install can spend at most 600 preparing plus its existing 600-second RPC envelope, excluding daemon startup. This is an explicit finite phase composition, not a claim that preparation already consumed the RPC envelope.

    3. For these artifact/source transfer paths, use a 30-second connection ceiling, a 30-second response-header wait, and a 120-second body-byte inactivity ceiling. For an upload, start the response-header wait after the outgoing request body finishes; do not charge a progressing upload body against a 30-second header timer. For a bodyless download request, the header wait follows connection establishment. Use backpressure-aware body progress; unrelated headers, keepalive traffic or timer callbacks are not body progress. Redirects and resume attempts may start fresh phase timers but never a fresh aggregate. Remove the contradictory fixed 120-second source-transfer and per-attempt 300-second upload ceilings. The earlier aggregate/caller deadline remains authoritative. Preserve an explicitly configured proxy upstream timeout as an independent operator cap; the current option is absent by default, not an implicit 30-second whole-transfer cap. Leave the dedicated health probe policy unchanged.

    4. Make Node ingress limits explicit at daemon and proxy constructors: 30 seconds for request headers and 600 seconds for receiving a request body, instead of relying on Node defaults. For upload transfer handling, start an independent 600-second aggregate at the request callback and enforce 120-second body inactivity. The incoming header phase has its separate ceiling. These limits do not impose a new 600-second cap on unrelated RPC handler execution or replace command deadlines. For artifact transfer forwarding, including Fetch embedding without a Node socket, impose a finite 600-second transfer aggregate and the same phase/progress policy. Propagate client disconnect and expiry through the existing request/stream owners. Abort a response body as well as the request when the downstream disappears.

    5. Keep resumable upload cleanup as a five-minute idle lifetime, not a five-minute absolute ticket expiry. Refresh it on actual received body-byte progress and after a successful appended chunk. Never refresh a transfer aggregate. A transfer may progress for more than five minutes only while both its aggregate and the ticket's refreshed idle lifetime remain valid. On inactivity or aggregate expiry, destroy the owned active streams and await existing rollback. A recoverable interrupted chunk returns to its prior verified offset. If rollback fails, invalidate the ticket. Legacy-upload failures remove their owned temporary directory. A later independently requested lifecycle may resume a still-live ticket under a new aggregate; an expired lifecycle itself may not resume or fall back. Keep authorization, tenant confinement, framing, byte caps, hashes, offsets and finalization semantics unchanged.

    6. Preserve typed caller cancellation. Produce typed timeout evidence from the owner of the expired aggregate or phase. Include the phase and effective budget without branching on error text. A timeout must not become a retryable transport error or be mislabeled caller cancellation. Keep existing cleanup and result-margin behavior. Do not add a transfer framework, cache, unbounded fallback or new retry policy.

    Source trace: daemon-client/index.ts:94-160 computes the RPC timeout, then connects and prepares artifacts before sendDaemonRequest. daemon-client-transport.ts:453-466 starts its monotonic RPC deadline at transport send. timeout-policy.ts:11-14,94-128 defines 480/600-second envelopes and the 30-second margin. daemon-command-registry.ts:49-54 derives the result deadline. HTTP http-server.ts:492-497 and socket transport.ts:91-94 register it. request-router.ts:230-234 applies descriptor execution deadlines. host-kit/internal/request-cancel.ts:221-258 owns deadline cancellation; its effective cutoff must also honor an earlier registered result deadline. server/platform-request-scope.ts:12-16 obtains that same request signal. platform-android/runtime.ts:452-456 passes it into deployment/runtime.ts:59-64, then the composition executor and install-artifact.ts:22-28 pass it to provision-kit materialization. The Apple deployment/executor path does the same. Native Android deployment/native.ts:8,25 and Apple core/config.ts:11 cap installation at 120 seconds. No 480-second native installer reservation exists in these owners.

    Source/body trace: provision-kit/install-source.ts:120-144 currently adds a fixed 120-second signal. install-source-download.ts:31-53 passes the signal through approved URL/redirect acquisition, but its writeResponse pipeline currently lacks the explicit signal. The installed Undici 7.30.1 api-request.js:62-70 destroys an already-created response body on abort; that source inspection is not runtime proof. Add the explicit pipeline cancellation and real post-header body-abort test. remote/upload-client.ts:62-112 starts preparation before its current per-request signals; upload-client-artifact.ts packaging/hash paths need cancellation. upload-stream.ts:173-205 owns the old 300-second timer and pipeline. Proxy upstreamSignal uses the caller signal by default and only adds a timeout when the operator configures one.

    Disconnect/cleanup trace: proxy/node-http.ts converts premature request/response close to the forwarded signal. Upload HTTP dispatch directly passes IncomingMessage into resumable receive, rather than using RPC lifetime registration. resumable-upload-transfer.ts:35-58 pipelines the real request to the file and rolls back to oldSize on failure; rollback failure invalidates the entry. resumable-upload.ts:183-209 owns idle expiry, active-receive destruction and queued directory cleanup. server/upload.ts removes its owned legacy temporary directory on failure. Preserve these owners and verify them through real loopback disconnects; RPC disconnect tests alone are insufficient.

    Illustrative policy timelines, not measurements: a 1 MiB source arriving in 2 seconds followed by a 20-second install remains well inside the existing envelope. A source that takes 330 seconds while delivering body bytes at least every 120 seconds, followed by a 120-second native install and 20 seconds of other work, consumes 470 of 570 seconds and can succeed; the old fixed download limit would fail it at 120 seconds. A 570-second work deadline still stops a continuously progressing download that leaves no installation time. A remote upload that needs 420 seconds with steady progress can complete inside its 600-second preparation aggregate, while a 120-second body stall or the 600-second absolute cutoff ends it. These choices do not promise success for every large or arbitrarily slow artifact.

    Implementation is staged by owner. First implement and author tests for remote preparation/upload aggregation and cancellation. Then implement source-install deadline/body propagation and explicit ingress/ticket inactivity behavior, with the shared policy values kept at their existing domain owners. Source workers must not modify proxy-origin work owned by #4434 concurrently. Before publication, use controlled clocks and real loopback client→proxy→daemon tests for progress beyond the old fixed limits, stalled headers/body, expiry despite progress, caller abort after response headers, two-hop disconnect, rollback/cleanup, ticket-idle refresh and no post-expiry fallback. Retain byte/framing/hash/redirect tests. Demonstrate old behavior or a meaningful policy mutant, perform independent review, and run exact-head validation with #4050 unchanged. No native device or cloud allocation is required. No merge, release or tags are authorized.

  5. thymikee commented on Nov 13, 2026

    @thymikee
    MemberAuthor

    Current state: The remote-only #4441 slice is committed and unpublished at f2eefda7da67a0c02b04e3729d17da3b1a9ad0d7 on base80075b6412dfa94fd765a36407270e21adce8981. The preserved preparation aggregate patch replayed unchanged. Pinned setup/build, typecheck, full format, diff/lint and143focused tests passed; no exact affected gate or whole-issue readiness is claimed.

    Independent source review found early-response cleanup/join and connection-phase ownership issues. The core findings are accepted for correction. The enclosing600-second aggregate still bounds the early-response case, so the report’s indefinite-hang wording is not adopted. Separately, the transform progress hook fires before a chunk reaches ClientRequest, so it cannot prove a socket has connected or made transport progress. A dependent OpenCode Apex medium task now corrects request/socket/body lifecycle and strengthens completed-transfer, early-response and typed-expiry proof. No publication until corrected source is reviewed and validated.

    The adopted finite policy is unchanged. Source-install cancellation, Node/proxy ingress and forwarding, ticket idle refresh, real two-hop disconnect/rollback proof, isolated regression proof and exact-head validation remain outstanding. #4050 stays selected. No timeout widening, blind CI retry, merge, release or tags.

  6. thymikee commented on Nov 13, 2026

    @thymikee
    MemberAuthor

    Decision: source-install deadline enforcement owner: Enforce the already adopted 570000ms source work budget through the existing descriptor/request-router path. Give both install-from-source and install_source one source-install timeout policy with600000ms client envelope and570000ms daemonDeadlineMs, derived from the existing30000ms result margin. The current withRequestDeadline owner must honor any earlier registered execution/result deadline, including refusing task entry when it is already expired. Do not arm every command’s result deadline globally or change unrelated command policies.

    The standalone materialization600000ms aggregate and incoming caller cancellation must be checked before publishing a completed source, even when owned work ignored an abort or resolved after its cutoff. Retain the existing platform native-install caps and request cancellation propagation. This selects the existing source-install descriptor boundary as the narrow enforcement mechanism; it does not change the finite budgets adopted in #4441 (comment). The real source-install request path, response-body cleanup, mutation and exact-head proof remain required before readiness. No global deadline framework, timeout widening, fallback, merge, release or tags.

  7. thymikee commented on Nov 13, 2026

    @thymikee
    MemberAuthor

    Current state: Two local components have bounded independent source reviews with no concrete findings. They remain separate unpublished commit chains and do not complete this issue.

    • Remote preparation/upload correction: fb7fbb8eee1cc4863256b1c38b32481b82900e7f. The review found the early-response abort/join and connection/body-phase corrections addressed within the remote slice. The worker reports 155 focused tests and targeted mutation detection. These supplied results are not an exact affected-gate pass.
    • Source-install/request-route component: d73159597a2e3a23b9b8db107437629a733dc8ef. Saved logs report 27 focused and 292 provider tests passing, plus typecheck/lint/layering. They used pnpm 11.25.0, so pinned 11.17.0 and exact-head validation remain required. The small optional-cause annotation correction is accepted. The worker's scope and shared-stash violations remain recorded; its candidate is clean, and the foreign stash entry was retained.

    Remaining: daemon/proxy ingress bounds, finite forwarding and downstream-disconnect propagation, ticket-idle refresh on body progress, rollback/cleanup, reconciliation of the deferred artifact-download path with the adopted policy, integration of the components, isolated regression/mutation evidence and exact validation. Preserve the adopted 600-second aggregates, phase ceilings, 570-second source work deadline and result margin. Progress must not reset aggregate deadlines.

    No PR or whole-issue readiness is claimed. Historical failed receipts and unproven timeout causes remain. No broad retries, timeout widening, test deletion, merge, release or tags.

  8. thymikee commented on Nov 13, 2026

    @thymikee
    MemberAuthor

    Current state: The preserved remote and source-install components are mechanically integrated at 5bad71760a5a2f41cd5198362c9ed5a5628ae478 on fresh main ddbfc85. Both ranges are patch-equal by git range-diff; the checkout is clean. Original components fb7fbb8eee and d73159597a and their evidence are preserved. This new head has not been validated.

    The next source-only slice refreshes the existing five-minute ticket idle lifetime during actual accepted body progress. The current code already refreshes after successful appended chunks. Node ingress, handler aggregates/body inactivity, proxy forwarding/disconnect, download phase coverage and full rollback/proof/exact validation remain separate obligations. External #4511 still overlaps deferred artifact-download work and is not maintained by this coordinator. The Apple4516 comparison owns local execution; source work does not run tests concurrently. No whole-policy completion or readiness claim.

  9. thymikee commented on Nov 13, 2026

    @thymikee
    MemberAuthor

    Current state: Ticket-idle review corrections are committed at29164a5cc593432414d3cc345aa1fdff93bf8784 on the integrated remote/source chain. Independent source review found no production regression and identified a proof gap. The coordinator corrected it: a stalled receive must remain pending before the refreshed expiry, then reject at the deadline with typed RESOURCE_EXPIRED evidence. The test no longer claims unasserted rollback. The empty-request case is retained as a closest-negative baseline control. The progress test now stays within the adopted ten-minute ingress envelope.

    Pinned focused tests passed12/12, typecheck, full formatting and diff-check passed. My first typed assertion overconstrained an additional hint and failed two cases; that test predicate was corrected and the failed log preserved. This is focused feedback, not an integrated exact gate or real socket proof. Next source-only slice makes daemon/proxy Node ingress limits explicit. Transfer aggregates, body inactivity, forwarding/disconnect, legacy cleanup, deferred download reconciliation and complete policy/regression/exact proof remain open.

  10. thymikee commented on Nov 13, 2026

    @thymikee
    MemberAuthor

    Current state: The Node ingress slice and its small review fixes are integrated into the remote/source/ticket candidate at3f88e9dee341031284b1a74629bd7ec229ae396b. Pinned format/build,24focused Node constructor tests,typecheck and diff-check passed. The daemon constant is private; the public proxyHTTPS options remain and have an embedding example. Tests assert constructed settings/public options, not live timeout expiry. Node periodic enforcement is not claimed as an exact wall-clock disconnect.

    The ticket slice separately passed12focused tests after the proof correction. No integrated exact gate or complete policy proof is claimed. Next source-only work adds the daemon streaming-upload600saggregate/120sbody-idle and disconnect/cleanup behavior at existing owners. Proxy forwarding/Fetch observability, download phase coverage, deferred artifact-download/#4511 reconciliation, real loopback and full regression/exact validation remain open. Reinstall's proof/gate worker alone owns local execution.

  11. thymikee commented on Nov 13, 2026

    @thymikee
    MemberAuthor

    Current state: The partial upload-lifetime review found a real error-precedence defect and premature idle timing while queued. Root corrected them directly at7bfce3773370e46c61917827de549a0637f26559. Typed rollback/finalize-cleanup/ticket-expiry errors are preserved; raw stream errors still map to the lifetime abort cause. Idle begins at the admitted pipeline, after the per-ticket queue. Three focused precedence cases passed. Typecheck initially found an authored unused import; after removing it,pinned typecheck/full format/diff-check passed. The failed typecheck log is kept.

    This is not full validation: optional-lifetime mode, six-case I/O/resource cleanup, real HTTP route proof, extraction cancellation, JSON-body idle and end-to-end phase classification remain open. The review's claim that legacy fallback gets a fresh600s client budget is false for this integrated code: one outer preparation signal is passed through direct,resume,finalize and legacy. Keep the explicit Node receipt limit and independent handler aggregate; adopted policy requires both.

    The proxy design confirmed Fetch visibility limits but is not adopted. A60s composite does not establish30sheaders after actual outgoing completion; consumed-body EOF can precede transport finish. A follow-up design must identify a supported transport owner and preserve the finite caps. No generic framework,private global hook,unbounded fallback or timeout weakening is authorized.

  12. thymikee commented on Nov 13, 2026

    @thymikee
    MemberAuthor

    Current state: Independent source-only review of the proposed default node:http(s) proxy adapter returned refine, then adopt. Before code, correct lifecycle listener registration and cleanup, preserve typed abort reasons, retain incomplete-upload308 without following redirects, preserve early responses without cancelling the client's successful incoming body, and keep the existing Request boundary/null-body status rules. Watchdog ownership, demand-pausing and runtime environment/Bun behavior need verification. Custom opaque transports cannot be falsely claimed joined after abort.

    The finite custom-transport/runtime/status decisions remain to be recorded. This design review is not a source implementation or whole-issue clearance. The current partial source candidate remains7bfce3773370e46c61917827de549a0637f26559, with daemon-lifetime proof/mandatory-owner cleanup, proxy implementation, end-to-end discrimination and exact validation still open. The adopted budgets are unchanged; no weaker fallback, fresh aggregate, new retry or arbitrary timeout was introduced.

  13. thymikee commented on Nov 14, 2026

    @thymikee
    MemberAuthor

    Decision: finite proxy transfer transport and observable error contract:

    Under the active completion authorization, adopt a proxy-owned node:http(s) default adapter for POST /upload, /upload/preflight, /upload/finalize, PUT /upload/direct/* and GET /artifacts*. Keep /rpc, health and diagnostics on their existing Fetch path. Preserve 30-second connect, 30-second headers after actual outgoing finish and connection establishment, 120-second body inactivity, and the non-refreshing 600-second forwarding aggregate. Caller cancellation and an explicitly configured operator cap still win. No retry, new aggregate after expiry, or Fetch fallback is added.

    Adopt the source-cited November14 transport plan with these binding corrections and finite choices:

    1. Register error/close/socket/drain/finish/response ownership synchronously at construction. Await only the close promises registered then; race drains against close and abort. Use the original aborted signal reason for classification. Destroy only an exchange's active owned work, never a shared global agent or a pooled socket after its ownership has ended.
    2. Never follow redirects; preserve the daemon's location-less308 and range/offset headers. Keep existing Request construction/URL validation and header allow-lists. Use urlToHttpOptions without auth. Accept final statuses200–599; HEAD and204/205/304 have a null body. Informational responses are not constructed as final Fetch Responses.
    3. Early401/413 and successful responses must reach the client intact. Stop the outgoing pump and handle later socket errors without replacing the accepted response. Do not cancel the incoming client source merely because a response was received. Any discard/drain remains bounded by the ORIGINAL aggregate/operator/caller lifetime, is joined or explicitly stopped before owned cleanup, and never outlives disposed timers. Do not await an endless client body before delivering the response. The implementation must prove this through real loopback cases.
    4. Use the runtime's default global Node agent, matching the existing remote client; do not create or destroy a private/global pool. Move the existing watchdog and outgoing-phase observation into host-kit/transport rather than copying them. Keep Node runtime loads lazy. Add suspend() only for demand-aware idle; dispose() still means settled. Remote behavior must remain unchanged by the move.
    5. Custom upstreamFetch keeps its signature. Bound proxy-owned settlement and enforce observable aggregate, operator/caller cancellation and demand-aware response/consumption idle. Do not claim a transport pull proves bytes sent. Document custom connect/header/stream/cancel obligations. Do not wait indefinitely for a foreign promise or cancel callback; safely handle late responses/rejections and request cancellation. No forced join of foreign work is promised.
    6. Keep existing HTTP statuses: INVALID_ARGS400, other failures500; no504. For recognized TRANSFER_PHASE_TIMEOUT_REASON or forwarding-aggregate errors only, expose validated top-level reason, transferPhase and finite timeoutMs. Do not add arbitrary details or change unrelated RPC/health error bodies. Record this additive wire change and test the actual HTTP response, even if the scanner does not require an acknowledgement.
    7. The forwarding aggregate has typed details {reason:'transfer_forwarding_deadline_exceeded',transferPhase:'forwarding',timeoutMs:600000}. Existing daemon phase/budgetMs spelling is not itself a defect; retain it unless an independently justified contract change is recorded. Integrity/rollback/finalize-cleanup failures must not be hidden by generic timeout causes.
    8. One default adapter serves supported runtimes; no runtime branch silently restores Fetch. Require Node floor and current-Node phase/lifetime proof. Require Bun compatibility evidence before asserting existing Bun support is preserved or claiming readiness; unavailable evidence is a draft blocker, not a documentation-only guarantee. Environment-proxy parity with Fetch is unproven and must be measured/disclosed.

    Explicit migration on DEFAULT TRANSFER routes: Location redirects are not followed; Fetch-injected default request headers are not sent; patched global fetch or an undici global dispatcher no longer intercepts that hop. Embedders needing that transport behavior supply upstreamFetch and own its documented phase guarantees. Environment routing follows node:http(s)'s global agent; no parity claim is made before proof. Document these changes with the owning proxy API, preserving TLS verification, authentication, tenant confinement, URL restrictions and framing.

    Start only source sliceA: the single phase owner/move and narrow suspend/observer work, reusing the preserved candidate. No proxy adapter or daemon expansion in that task. Apple4442 retains the sole test slot; new tests are authored-unrun. The proposed A/B/C names are source boundaries, not proven independent PR layers: the watchdog exists only in unpublished4441 remote changes. Before publication, include and validate each required baseline/dependency using gh-stack, preserve a self-contained tested layer, and respect gross-diff limits. Do not split detecting tests away merely to shrink an implementation PR. Split source by independently enforceable contracts, not by hiding proof.

    Proxy implementation follows review of sliceA and this adopted contract. Daemon gaps and downloads retain separate owners; external4511/4505 are dependencies/overlap only and are never maintained. Required evidence includes real-loopback308, early401/413, close/drain races, two-hop disconnect/rollback, ignored custom signals, backpressure, aggregate despite progress, late opaque settlement, Node/Bun/environment routing, wire/closure checks and meaningful mutants. Source reading is not runtime proof. Preserve4050, historical failed receipts and unproven causes. No merge, release or tags.

  14. 2 remaining items

  15. thymikee commented on Nov 14, 2026

    @thymikee
    MemberAuthor

    Current state — completed-hop retirement review and next proof

    The bounded source review found no defects in the lower 6089d90 watchdog retirement or the upper 43abdfa8 observer retirement. This was source reading only; the resumed-upload regression has not yet been demonstrated at runtime.

    One Claude owner is now adding the lower-owned loopback regression test, restacking the existing upper layer with gh-stack, and running focused checks in sequence. The test must preserve a retryable peer reset after a resumed hop makes real progress beyond the completed hop’s old 120-second ceiling. It will use the existing parked-quota fixture.

    Mutation experiments will use separate ownership after the heads are frozen. Each layer still needs its own exact-SHA gate, wire compatibility evidence and relevant closure/export checks. Observer listener removal is unit-testable; its private upload call site is source-verified, so no runtime discrimination is claimed for a behavior-equivalent mutant. The stack metadata discrepancy must be resolved without dropping the actual upper commit.

    No proxy, daemon or download implementation is included in this task. No PR is published for these layers, and no device or external PR work is assigned.

  16. thymikee commented on Nov 14, 2026

    @thymikee
    MemberAuthor

    Current state — resumed-hop proof test corrected; independent checks assigned

    The first test was unstable: hop two’s own request-body idle budget expired despite peer byte growth. Those failed runs remain recorded, and no regression proof or exact gate is claimed from them. A quota-only fixture change also failed and was reverted.

    The corrected test observes the real node:http request through a pass-through spy. Each virtual window requires both new peer bytes and a new drain event from the resumed client request—the event the watchdog actually consumes. It uses a 32-MiB fixture to keep the body incomplete without changing any timeout. The final focused runs passed 54 tests on lower 9544701 and 65 on upper 9d5a56942e. Lower typecheck/lint, full formatting and the upper build passed. These are focused results, not exact affected-gate receipts.

    The upper layer was restacked with gh-stack. Its observer-retirement patch is retained, and both metadata heads now match the actual refs. Both layers are clean and unpublished. One Claude reviewer is reading the frozen changes; a separate sole execution owner is testing old-code discrimination in an isolated checkout, lower then upper. Separate exact-head gates, wire/closure/export evidence and full policy completion remain open. No proxy, daemon, download, device or external PR work was added.

  17. thymikee commented on Nov 14, 2026

    @thymikee
    MemberAuthor

    Current state — bounded test review complete

    The independent source review found no blocking defect in the drain-gated resumed-upload test at lower 9544701 / upper 9d5a56942e. It confirmed that the spy observes the real transport boundary and that each progress window waits for the event the production watchdog consumes. This is source reasoning, not old-code detection.

    One low finding corrects an inaccurate comment about the first answered body remaining unread; one nit decouples the test’s progress window from the connect-budget constant. Root prepared those small changes but has not applied them. Both protected heads remain frozen while the sole isolated execution owner finishes R1/M-U1 and the behavior-equivalent observer control. The current evidence will retain its original SHA attribution. No extra execution owner or review worker was started.

  18. thymikee commented on Nov 14, 2026

    @thymikee
    MemberAuthor

    Current state — hop regression detected; structural checks failed

    At lower 9544701, the faithful old production hunk made the new resumed-hop case fail: 1 failed, 53 passed. Restoring the fix returned 54 passes. At upper 9d5a56942e, removing only the completed-hop watchdog disposal caused the same intended phase-classification failure: 1 failed, 64 passed, then 65 passed after restoration. Removing only observer disposal survived, as predicted; that call site is not runtime-detected. These are isolated patched-tree experiments, not exact affected-gate receipts. The logs showed a phase timeout but did not capture its exact phase and budget.

    The structural preflight found real failures: eager-closure growth and stale wire acknowledgement at both layers, plus four unused public host-kit re-exports and two R84 test-colocation names at upper. Those failures are preserved; no threshold, allowlist, test or gate was weakened.

    Root applied the small comment/window review fixes on lower 7e03b3e and restacked upper 3595691b. The 54 lower focused tests still pass; the assertion now prints error details for the next isolated run. One Claude owner is repairing the closure/module boundaries, public exports, test names and per-declaration wire acknowledgement across the existing stack. Budgets, admission, cancellation and retry policy remain fixed. New frozen-head review and separate exact gates will follow; no full gate has run and no layer is published.

  19. thymikee commented on Nov 14, 2026

    @thymikee
    MemberAuthor

    Current state — structural repairs pass; exact validation assigned

    The existing transfer stack is clean and unpublished at lower 09e1e76 and upper 217528a. Both layers now pass the closure pair (918 tests), wire tests (50) and released-wire check, layering, production exports and Fallow. The wider focused sets passed 86 tests at lower and 103 at upper; the original control counts remain 54 and 65. Formatting, lint and builds/typecheck passed, with upper build/typecheck attributed to its identical production source before the final docs-only commit. No exact affected gate has run yet.

    The lower helpers now live in the already-loaded upload-stream owner; the upper phase owner and observer live in host-kit’s already-loaded transport internals. Superseded modules are deleted. Admission stays synchronous, and budgets, cancellation, retry rules and wire bytes are unchanged. Test renames retain all assertions. Only the actual changed declaration’s wire digest and acknowledgement were updated per layer. The new redundant agent-doc line was removed to satisfy its byte budget. These choices are explicitly included in the fresh source review.

    One Claude reviewer is reading the frozen heads. One separate execution owner is rerunning isolated lower/upper discrimination and then one exact affected gate per layer in sequence, with the full-unit fallback preserved. The previous failures and proof remain at their original SHAs. Node-floor, Bun, proxy, daemon and remaining transfer-policy work are not established by this slice. No PR is published or merge authorized.

  20. thymikee commented on Nov 14, 2026

    @thymikee
    MemberAuthor

    Current state — two transfer layers published, review feedback owned

    The existing #4441 remote-upload layer is published in #4552 at 09e1e76. Its shared phase-owner layer is #4551 at 217528a, based on the lower branch. Both are ready for review in stack #4553; neither is merged.

    Each head passed a separate exact affected gate: nine passed checks, seven CI-owned/parked skips, exactCommit true, checkout unchanged and no fail-open. Full unit passed 19,818 tests at lower and 19,829 at upper, ten skipped in each; related passed 376 in each. Isolated old-code/watchdog mutants failed at the intended response-body/120000 error, and restored controls passed 54/65. Observer disposal remains source-verified with unit-tested listener removal; its behavior-equivalent mutant survives. Historical failed receipts remain intact.

    GitHub CI and reviews are now authoritative next steps. Three small #4551 test/comment requests are accepted. One additional request questions drain-only progress for successful small writes; a bounded Claude-owned real loopback will verify it before any correction. The 30/30/120/600 policy is unchanged. The initial size reports show no budget violation.

    These layers do not complete the proxy, daemon, source-install or download portions of #4441. The canonical issue stays open. No cleanup, external PR maintenance, merge, release or tags were performed.

  21. thymikee commented on Nov 14, 2026

    @thymikee
    MemberAuthor

    Current state — published transfer feedback and measured write progress

    Both published layers remain open and unmerged at #4552 09e1e76 / #4551 217528a. New local review corrections are lower fae463f and upper 5772b18bc8 (not pushed or exact-gate validated).

    The coordinator added optional onChunk observation to host-kit's existing response reader and deleted the upload copy. It fixed TLS/pool fixture cleanup, replaced an exception in an HTTP request listener with a 500 response, pinned remote phase expectations as literals, and pinned zero timers after an already-aborted arm. A canceled retry now rethrows its own signal before emitting fallback. The unused suspend surface and its five exclusive tests are deferred together to the proxy consumer; the eight original watchdog tests and six observer cases remain. Focused controls pass 31 lower and 37 upper tests, with full formatting and typecheck. One transient duplicate-import error during replay was corrected; its failed log is retained. These are focused results, not new exact receipts.

    The real HTTP spike confirmed the drain-only observer defect: 1 KiB successful writes produced zero drains, yet the request-body watchdog expired at 120 seconds after four completed writes and 4096 received bytes. The default 64 KiB CLI path did not reproduce on Node 26.11.1. Successful write completion is the adopted progress boundary: bytes handed to the local kernel, not peer receipt or acknowledgement. No Transform/read-ahead, queued write, empty write, failed callback or post-response/abort/dispose callback may refresh the body phase. The 30/30/120/600 policy and retry rules are unchanged.

    One Claude source owner will repair that boundary across the independently valid lower layer and the shared upper owner, and repair the small-artifact header-wait regression proof. Fresh independent review and separate exact gates precede one review-round push. The existing aggregate admission boundaries must still cover standalone uploadArtifact calls; only demonstrably redundant attempt-level timing may be removed. The optional broader aggregate/stack reshaping is deferred rather than silently changing standalone admission or increasing coupling.

    Coverage's original logs end without a test-failure summary, so their causes remain unproven. Per the explicit human request, each failed Coverage job was rerun once, with the original logs retained. Results and moved-test collection will be captured; no repeated retry or passing claim is made. The two new PR watches follow reconciliation and the completed review-round push; #4544's existing watch is unchanged. No merge, release, tags, foreign resources or queued implementation.

  22. thymikee commented on Nov 14, 2026

    @thymikee
    MemberAuthor

    Current state — transfer feedback source frozen; Linux Coverage residual remains

    The existing #4552/#4551 source correction is committed locally at lower c9f91d0 and upper 5687f46cefdcace44d8ccccf052e8f7b1edcf9c5. The stack is clean and gh-stack metadata matches. No push or exact-head gate has occurred. Author-focused controls passed 65 lower/74 upper tests; structural, closure, wire and provider-focused checks passed. Those results do not replace GitHub Coverage.

    The write-completion correction, shared-reader reuse and stronger header-wait proof retain the finite budgets. The redundant attempt aggregate/hints are removed after a caller audit; standalone uploadArtifact and request preparation remain bounded. Source-author mutation runs modified the implementation tree and were restored; they are patch-input feedback, not the required isolated proof. Fresh isolated discrimination remains owed. Lower gross size is 2491 lines, upper 1039, so the independent review explicitly includes a smaller design.

    All six upper/three lower annotated Linux Coverage failures remain unresolved; no matching Linux reproduction or causal diagnosis is claimed. Node24/macOS and local focused V8 passes do not establish Linux behavior. No local Linux runtime/container or linked-environment tool is available. One read-only Claude reviewer examines the frozen heads, and one separate execution owner studies a concrete controlled real-I/O scheduling counterfactual in the existing isolated experiment. That is modeled evidence, not native Linux proof. No new CI retry, timeout increase, broad gate, device action, merge, release or tag.

  23. thymikee commented on Nov 14, 2026

    @thymikee
    MemberAuthor

    Current state — transfer source review complete; fixture and test-power findings open

    The independent source-only review accepts production at lower c9f91d0 / upper 5687f46c, but does not clear publication. Its open findings cover fake-clock/real-I/O synchronization and bare waits, lower-layer negative-test power, and unfinished/backpressure premises. The Linux Coverage failures remain failed with cause unproven; neither source reasoning nor macOS passes establish Linux behavior.

    The existing isolated execution owner received these findings and continues the already-assigned fixture study. The protected heads stay frozen. No second execution owner or gate was started. The coordinator will handle the small encoded-empty-string correction after the isolated owner finishes, preserving valid string progress and Node callback semantics. Final test coverage, size boundaries and exact-head receipts remain owed. No source edit, push, CI retry, merge, release or tag occurred in this reconciliation.

  24. thymikee commented on Nov 14, 2026

    @thymikee
    MemberAuthor

    Decision: first-layer completed-write ownership and faithful fixture timing:

    The isolated macOS delayed-fs.read experiment demonstrates a fixture defect, not the native Linux CI cause. Adopt its event-before-fake-time direction, with per-test waiter cleanup and explicit settlement/premise assertions. The existing test timeout bounds real I/O; do not add arbitrary sleeps, widen timeouts/loop limits, or claim the Linux failures fixed before new-head CI.

    Do not accept the lower-layer test gap confirmed by six surviving private-observer mutants. Move the existing completed-request-write observer into host-kit's already-loaded transport owner in the first layer, with its generic contract tests and its actual remote production consumer. The callback reports successful nonempty encoded bytes handed to the local kernel, never peer receipt. This concrete consumed API replaces the private remote copy; do not introduce another module, framework or test-only export.

    The upper phase-controller move remains a separate independently valid layer. It must reuse the same write owner, with one write-progress observer per request and separate phase-event ownership. Keep the completed-write API genuinely consumed at each layer; no unused public shim. Put direct callback/encoding/error/response/abort/destroy/dispose/restoration/foreign-wrapper tests with the first owner. Retain consumer phase and lifecycle evidence. This is a partial host-kit-first correction, not a duplicate implementation or a new audit scope.

    The coordinator authored the small encoded-string correction at upper bdc9452100: valid string progress stays counted; encoded-empty hex/base64/base64url cannot refresh inactivity. Its 21 focused owner tests, full format and typecheck pass. The first-layer owner must carry the same correction and tests before publication; the former private lower copy is superseded.

    Replace the fixture's host-speed rounds<20 condition only after an isolated counterfactual proves a retained, stronger owned-destruction/close/join invariant detects a real unjoined pipeline. Do not delete a detecting assertion merely to make delayed I/O green. Assert that stall/join data is still unfinished rather than silently relying on buffer-size guesses.

    Keep 30/30/120/600, standalone admission, retries, cancellation, ticket idle and wire semantics unchanged. Apply the repository's size bounds conservatively to actual PR ranges and the main-to-upper cumulative range; do not raise limits or discard detecting tests. Remove duplicate narrative comments only when code, names or public constraints already express their meaning. Fresh source/smaller-design review, isolated discrimination and separate exact-head gates precede one review-round push. No merge, release or tags.

  25. thymikee commented on Nov 14, 2026

    @thymikee
    MemberAuthor

    Current state — first-layer write-owner correction complete; final review and exact gates assigned

    Local lower #4552 is ad4603b and upper #4551 is 4a6dc5ff7fad7f788e82182fc38c6dae86b09378, clean and unpushed. The completed-write owner and direct contracts now land with the first real consumer; no private remote copy remains. All six formerly surviving lower negatives are detected in the isolated experiment. The encoded-string correction is retained. Modeled macOS delayed-read controls pass 18/18 at D0/D1/D3 on both heads; this is not native Linux evidence.

    The event-driven fixtures include test-lifecycle cleanup and stronger destroy-before-next-open checks in place of host-speed or kernel-buffer-dependent assertions. Stall tests now check unfinished-body premises; Linux buffer capacity remains unmeasured. One no-destroy mutant exposes the intended hang through the 5000 ms test limit; the other reported kills are assertions. No blanket deletion or timeout change was used.

    Author-focused results: 76 lower/82 upper, wire50, closure969 and provider6 on each. Structural checks and full formatting pass. All size ranges are below2500: lower2326, upper741, cumulative2471. No exact gate has yet run at these heads. A fresh read-only source/smaller-design reviewer and one isolated exact-gate owner now validate them; lower then upper, with the full-unit fallback preserved. Sources stay frozen.

    Published heads09e1e760/217528ae and their original/rerun Coverage failures retain their attribution. No new CI retry, push or readiness claim is made. New-head GitHub Coverage after the reviewed/gated push remains authoritative. No merge, release, tag or device action.

  26. thymikee commented on Nov 14, 2026

    @thymikee
    MemberAuthor

    Current state — remaining review corrections committed; new exact heads assigned

    Both previous ad4603b/4a6dc5ff heads passed separate local exact gates (19,831/19,837 unit tests passed, ten skipped each, no checkout change or fail-open). Those receipts retain their original SHAs. Review accepted production and identified small test-fidelity corrections, now committed locally at lower22cf1d4993ffbf7d7f29f8dd4825e64b6a5f52ab and upper10ea602e701059acfb88140468e95b1cefd5d3c7.

    The tests now wait for actual response data, outgoing finish, a real redirected request or body progress before charging time/aborting. Held-open upload sources make unfinished-body premises structural; names do not claim peer-induced backpressure. Hashing remains real. Duplicate silent TCP fixtures and a vacuous paused-server counter are replaced by one origin and a client-write absence pin. Uppercase encoding and realm-safe Uint8Array cases are covered. Root-focused tests passed30lower/44upper; typecheck, lint, full format and closure/wire checks passed. Size ranges2337/741/2482remain below2500.

    One strict read-only recheck and one isolated proof/exact-gate owner now validate the new heads, sequential lower then upper. No push yet; old published Coverage failures remain failed, with Linux cause unproven. New-head GitHub Coverage remains authoritative.

    For the ticket question: current daemon code refreshes its five-minute timer after an appended chunk completes, not on each in-flight byte. The adopted per-byte refresh remains separate daemon work under#4441. The client's600-second aggregate does not renew that timer; this remote-only stack makes no end-to-end long-chunk guarantee. No daemon edit, CI retry, merge, release or tag.

  27. thymikee commented on Nov 14, 2026

    @thymikee
    MemberAuthor

    Current state: corrected transfer review round published; exact local gates passed, Linux CI pending

    #4552 is now at 22cf1d4 (base main); #4551 is at 10ea602 (base the lower branch). Both were pushed together once after the fresh source recheck found no actionable findings and the separate final exact-head gates passed. Sizes are 2,337 / 741 / 2,482 gross lines for lower / upper / cumulative.

    Each pinned affected gate passed nine checks and skipped seven GitHub/manual lanes, with exactCommit=true, checkoutChanged=false, failOpen=false and the full-unit fallback unchanged. Lower unit: 19,832 passed/10 skipped; upper: 19,838/10; related: 376 each. Three new isolated counterfactuals detect lost uppercase encoding normalization, realm-safe byte detection and the adopted 120 s inactivity budget. Historical receipts retain their original SHAs.

    The shared response reader, first-layer completed-write owner and direct negative tests, faithful I/O-before-clock fixtures, structural unfinished-body premises, cancellation telemetry and unused suspension API corrections are published. Eleven existing inline threads have replies and dispositions. Human-review responses: #4552 (comment) and #4551 (comment).

    Old-head Coverage failures and their single captured reruns remain failed; their Linux cause is unproven. Corrected-head Linux/Coverage and new review feedback remain pending. The current daemon refreshes its five-minute ticket timer only after completed chunk append; adopted per-byte refresh remains separate daemon work. No proxy, daemon, download, Node-floor, Bun or whole-#4441 completion is claimed.

    The local execution owner is released. Initial watches will follow the final missed-comment refresh. The single published-only backstop remains enabled; no queued scope is started, cleanup performed, merge made, release made or tag pushed.

    Post-push update: upper-layer Coverage passed on 10ea602 (job 132549328094). Lower-layer Coverage failed on 22cf1d4 (job 132549325094): the header-budget, reused-connection and zero-length-artifact tests each hit 5000 ms. No retry was made and the cause remains unproven. A new Cubic control-plane timeout/fallback finding is accepted and remains open. One Claude read-only diagnosis owns the captured CI failures; no local heavy-test slot is active. Initial transfer watches remain deferred until this new feedback is processed and any required branch update is complete.

  28. thymikee commented on Nov 14, 2026

    @thymikee
    MemberAuthor

    Current state: both transfer layers CI green and handed back for human review

    Both published #4441 transfer layers are CI green and handed back for human review: #4552 at 13a4c1d9576777989ab3849cda9fb78661a33988, dependent #4551 at 024cb04a4d38de644148233ae79108191a28507d. No review threads are open. Both dedicated watches were removed after processing existing feedback. No merge or label action was taken.

    Lower Linux Coverage job132601152802 passed20,190 tests (70 skipped), plus11 fuzz tests. Upper Linux Coverage job132601156918 passed20,196 tests (70 skipped), plus11 fuzz tests. Integration and current Apple smoke lanes passed on both heads. The upper's old failed aggregate belongs to cancelled run43645155990; replacement run43645156143 passed without a retry. Sentinel approves both incremental reviews. The lower mentions an undisclosed off-diff candidate; the upper filters two below-minor items. No actionable payload was exposed for those notes, so they remain review limitations, not full-repository clearance.

    The existing public replies carry exact-head local and CI evidence: #4552 (comment) and #4551 (comment). Fresh cold review found no findings. Separate exact local gates passed with exactCommit=true, checkoutChanged=false and failOpen=false: nine checks passed, seven CI-owned/manual checks skipped per head. The #4050 full-unit fallback passed19,841 lower/19,847 upper tests,10 skipped each;376 related tests each. Focused baselines passed121/127. Four counterfactuals failed at their intended assertions; restores were byte/hash identical. Gross sizes remain2404 lower,699 upper,2497 combined.

    The lower header phase starts at connected request finish; pipeline cleanup cannot delay or reset it. The upper retains one phase owner and joins before outcomes/next hops. Historical failed receipts retain their original SHA/status; actual old CI close latency remains inferred. The separate process-fixture correction remains canonical in #4544 and was not duplicated.

    This completes this published feedback round, not whole #4441. The adopted30/30/120/600 policy, caller precedence and no-expiry fallback remain binding. Proxy/daemon/download/source-install obligations, per-byte ticket refresh and Node-floor/Bun evidence remain separate. #4544 stays at human handoff with its native-code60/memo limitations. Other canonical scopes retain their Decision prerequisites and sequencing. The local execution owner is released. The one published-only backstop remains enabled until all authorized scopes settle. No queued implementation, cleanup wave, new timer, release or tags.

  29. thymikee commented on Nov 15, 2026

    @thymikee
    MemberAuthor

    Current state: both published transfer layers merged by the maintainer; transfer scope remains open

    The maintainer merged #4552 at 2026-11-15 07:52:38 UTC (reviewed head 13a4c1d9576777989ab3849cda9fb78661a33988, merge commit d66323e8eaef83c2fb46ffe55fd6f2b09bb00754) and #4551 at 07:52:40 UTC (reviewed head 024cb04a4d38de644148233ae79108191a28507d, merge commit d64ac728b69185493af1b3b6b61dd3a79ba41814). GitHub comparisons confirm that main contains both merge commits. The coordinator observed these merges and did not perform them.

    Current-head Coverage, Integration and Apple evidence, independent reviews, regression controls, exact local receipts and historical failures retain their original SHA attribution. The old cancelled upper aggregate is not a new failure or a reason to retry CI. The lower off-diff and upper filtered review limitations remain recorded.

    This completes the two published remote/shared-owner layers, not all of #4441. Proxy, daemon ingress/lifetimes/ticket refresh, source-install/download propagation and remaining runtime evidence retain their existing finite Decisions and owners. No queued implementation or cleanup starts from this backstop. Both dedicated watches were already removed for human handoff. Retain shared resources and durable proof. The single published-only backstop remains enabled. No release or tags.

    Authorized cleanup (Chief of Staff, 2026-11-15): removed 11 clean, unused worktrees and 16 obsolete local branches for #4544/#4551/#4552. A verified Git bundle preserves their branch history with the existing durable audit evidence. The three remote PR head branches were already absent. Only the merged transfer stack's local tracking was removed; GitHub was not changed.

    The shared coordinator and four candidate worktrees for unfinished #4441 transport/source-install/ingress work remain intact. Dirty and unrelated resources were not cleanup targets. T3 device inventory confirms the recorded Apple simulator is stopped and this thread has no open device panels. Both recorded private daemons and runtime directories were already gone. No foreign device, session or daemon was stopped. No implementation, test/gate, merge, release, tag or timer operation was performed. The single backstop remains enabled for the six unfinished authorized scopes.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    enhancementNew feature or request

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions