test(continuation): trap mode-only PendingContinuationDelegate at compat boundary by silas-dandelion-cult · Pull Request #462 · karmaterminal/openclaw

silas-dandelion-cult · 2026-04-30T06:10:08Z

Closes #438.

What this PR adds

A single structural trap test at src/auto-reply/continuation/types.mode-shape.test.ts (228 lines, test-only — no production code changes) that pins three load-bearing invariants:

Runtime objects are mode-only. consumePendingDelegates and consumeStagedPostCompactionDelegates return PendingContinuationDelegate shapes whose only mode-bearing field is mode. They MUST NOT expose silent / silentWake / postCompaction boolean runtime fields. Asserted across normal / silent / silent-wake / post-compaction modes (4 cases via it.each).
Tool descriptor exposes mode enum, not boolean flags. continue_delegate parameter schema advertises mode as an enum over [normal, silent, silent-wake, post-compaction] and exposes NO silent / silentWake boolean parameters. (Schema-rendering robust: parses both enum and anyOf forms emitted by optionalStringEnum.)
On-disk TaskFlow stateJson MAY still contain legacy booleans. Positive back-compat assertion: persisted stateJson projects mode='silent' → silent: true etc. This is what justifies the runtime/disk encoding split. Asserted across the three non-normal modes + a clean normal row.

Why this is the trap that catches the next #433

#433 broke push because the squash silently reintroduced double-encoding (booleans + mode) at the compat boundary. The behavior tests didn't catch it because they assert behavior (does silent-wake wake?), not shape (does the runtime object expose only mode?). This trap asserts the shape, full stop.

Comments in types.ts and delegate-store.ts (#227 ADR) explicitly call out the encoding split as load-bearing, but until now nothing rejected boolean-shaped runtime delegates structurally. This PR closes that gap.

Verified load-bearing

Temporarily added ...(state.silent === true ? { silent: true } : {}) to flowToDelegate in delegate-store.ts — i.e., the exact regression class #227 warns against. Result: 1 of the 9 trap tests failed with the canonical runtime PendingContinuationDelegate must not expose 'silent' (mode-only encoding) message. Reverted. 9/9 passing on canonical2 cf7830ffb3702bf7d826d70838893e2e41709f12 baseline.

Test surfaces

7 runtime-object cases (4 mode variants × consumePending + 1 staged-post-compaction + 2 disk-back-compat)
1 tool-descriptor case (mode enum present + boolean params absent)
1 disk-back-compat negative (normal mode → no boolean flags persisted)

Receipts

Test Files  1 passed (1)
Tests  9 passed (9)

Refs

[architectural-decision] Pin mode-only PendingContinuationDelegate at compat boundary #438 — issue (frond-scribe walker, Project 56)
step-9 squash type-inconsistency: legacy delegate-store + missing ChainState import (incomplete substrate fold) #433 — broken push that triggered the walker walk
Type: PendingContinuationDelegate — drop redundant boolean encoding (use mode only) #227 — mode-only ADR (the canonical seam this trap defends)
canonical2 base SHA: cf7830ffb3702bf7d826d70838893e2e41709f12

🌫️

…pat boundary Adds structural trap test that pins: 1. Runtime objects from consumePendingDelegates and consumeStagedPostCompactionDelegates expose only `mode` — no `silent` / `silentWake` / `postCompaction` boolean runtime fields (mode-only encoding per #227). 2. continue_delegate tool descriptor advertises mode as an enum over [normal, silent, silent-wake, post-compaction] and exposes no `silent` / `silentWake` boolean parameters. 3. On-disk TaskFlow stateJson MAY still project to legacy boolean flags — back-compat positive assertion that justifies the runtime/disk encoding split. Verified load-bearing: temporarily leaking `silent: true` from flowToDelegate fails the runtime-object suite (1/9 fail), proving the trap catches the regression class #433 nearly reintroduced. 9/9 passing on canonical2 cf7830f baseline. Closes #438

chatgpt-codex-connector · 2026-04-30T06:15:25Z

💡 Codex Review

openclaw/src/auto-reply/reply/agent-runner.ts

Lines 2738 to 2739 in 0d0a93a

    
           if (delegate.delayMs && delegate.delayMs > 0) { 
        
             const clampedDelay = Math.max(minDelayMs, Math.min(maxDelayMs, delegate.delayMs));

Stop re-delaying delegates already matured by store consume

This branch applies a fresh setTimeout whenever delegate.delayMs > 0, but these delegates were just returned by consumePendingDelegates(...) in the same block, which only yields entries whose original delay horizon has already matured. Reapplying the delay here causes delayed delegates to wait an extra full interval (and after a restart, can postpone work far beyond the configured delay), which breaks expected continuation timing for any non-zero delaySeconds.

openclaw/src/agents/tools/continue-delegate-tool.ts

Lines 108 to 110 in 0d0a93a

    
           const currentCount = 
        
             pendingDelegateCount(sessionKey) + stagedPostCompactionDelegateCount(sessionKey); 
        
           if (currentCount >= maxPerTurn) {

Enforce maxDelegatesPerTurn using turn-local count

The limit check uses pendingDelegateCount + stagedPostCompactionDelegateCount, which counts all currently queued delegates for the session, including backlog from earlier turns. That means a user can hit "maxDelegatesPerTurn" on a new turn solely because an older delayed/post-compaction delegate is still queued, even though this turn has not exceeded its own budget.

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

elliott-dandelion-cult

🌻 — APPROVE. Test-only trap (+228/-0 single new file). Three load-bearing assertions:

Runtime objects from consumePendingDelegates / consumeStagedPostCompactionDelegates expose only mode — no silent/silentWake/postCompaction boolean runtime fields.
Tool descriptor advertises mode as enum [normal, silent, silent-wake, post-compaction], no boolean parameters.
On-disk TaskFlow stateJson MAY still project legacy boolean flags — positive assertion preserving runtime/disk encoding split.

Verified-load-bearing (temp silent: true leak from flowToDelegate fails 1/9). Catches the regression class #433 nearly reintroduced. 9/9 passing on canonical2 cf7830f.

🌻

…getSessionKey absence (#463) Adds [#446] complementary trap to the [#438] mode-only trap (PR #462): - Closed-set assertion on tool.parameters.properties keys: exactly [task, delaySeconds, mode], no more, no less. Catches additions of new model-facing parameters (cross-session addressing, retry knobs, priority) without an ADR. - mode enum membership pin (normal, silent, silent-wake, post-compaction) — robust to typebox enum vs anyOf rendering. - Boolean-runtime compatibility fields (silent, silentWake, postCompaction) MUST be absent at the descriptor surface. On-disk back-compat lives in the Zod state schema (#438), not here. - targetSessionKey absence reaffirmed with load-bearing reason in the failure message (cross-session addressing is the (b)-shape lane, binary-canticle#11). Verified load-bearing: temporarily added targetSessionKey to ContinueDelegateToolSchema, the closed-set assertion failed with the expected got=[…, targetSessionKey, …] message. Restored. 9/9 passing on canonical2 cf7830f baseline. Closes #446

Three test files from merged PRs (#462, #468, #511) were absent because this branch forked from canonical2 before those PRs landed. The post-revert allow-list audit (§3.4) flagged them as deletions from landed PRs. Restored from canonical2 HEAD (74940e5). - types.mode-shape.test.ts (#462) - agent-runner.continuation-span-uniformity.test.ts (#511) - store.continuation-merge.test.ts (#468) tmp-drop-me-otel-span-uniformity.md omitted (copilot scratch; safe to drop). Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

…6.4.24) (#515) * wo(canonical2-rebase-pathB): rebase Path-B's 5 cleanup commits onto canonical2 (figs directive 22:55Z) * chore(v3-cleanup): wave A cohort-identity scrub Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * chore(v3-cleanup): drop rejected rebase artifacts Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * docs: scrub workspace template wording Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * refactor(v3-cleanup): wave B structural dedup of continuation runtime Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * docs: journal canonical2 wave B Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * fix(v3-cleanup): wave C import discipline and build warnings Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * docs: journal canonical2 wave C Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * fix(v3-cleanup): wave D surface continuation failures Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * fix: surface compaction count reconcile failures Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * test(v3-cleanup): wave E continuation coverage Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * docs: journal canonical2 wave E Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * fix: align bundled plugin dependency types Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * test: isolate bedrock app profile runtime deps Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * chore: scrub fork process labels from source comments Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * fix: close continuation type design blockers Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * chore: scrub continuation prompt process link Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * docs: journal canonical2 final checkpoint Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * Revert "chore(v3-cleanup): drop rejected rebase artifacts" This reverts commit 3396b88. The original commit mass-deleted 30 files (6745 deletions) under the label "rejected rebase artifacts." ~5141 of those deletions are landed swim-37 durability harness substrate from merged PRs #412/#413/#414/#416/#417/#418/#419 plus collateral docs/scripts. These are not rejected artifacts — they are committed, merged test infrastructure that proves continuation durability across compaction. Cohort review (🩸 + 🌊 + 🌻 + 🌫) confirmed the block finding at PR #515 issuecomment-4362337067. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * docs: release-note for context-pressure band-derivation behavior change Wave B (cefa09d) changed context-pressure bands from fixed [25, 80, 90, 95] to threshold-derived [thresholdPct, 90, 95]. At default 0.8 the implicit 25% early-warning band is removed. Ship-acceptable per cohort review; release-note documents the change and points to #516 for the earlyWarningBand config opt follow-up. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * fix: restore landed-PR tests missing from rebase fork-point Three test files from merged PRs (#462, #468, #511) were absent because this branch forked from canonical2 before those PRs landed. The post-revert allow-list audit (§3.4) flagged them as deletions from landed PRs. Restored from canonical2 HEAD (74940e5). - types.mode-shape.test.ts (#462) - agent-runner.continuation-span-uniformity.test.ts (#511) - store.continuation-merge.test.ts (#468) tmp-drop-me-otel-span-uniformity.md omitted (copilot scratch; safe to drop). Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * fix: add rebase.classify to ContinuationSpanName for restored tracer The revert of 3396b88 restored src/rebase/tracer.ts which emits "rebase.classify" spans. Commit 4871c81 (fix: close continuation type design blockers) narrowed startSpan from string to ContinuationSpanName after tracer.ts was deleted — additive fix to include the span name in the union. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * feat(continuation): add earlyWarningBand config opt for post-compaction cycle primer * test(continuation): pin earlyWarningBand default-preservation + opt-out branches * fix(continuation): add curly braces to satisfy linter * fix(continuation): unblock early-warning band fire path + make field optional Three bugs caught in cohort review of v5 (3e88ce5): 1. Suppression guard bug (Silas): non-postCompaction call sites bailed with 'ratio < threshold' BEFORE the resolved early-warn band could fire. Even with earlyWarningBand explicitly set, ratio=0.25 + threshold=0.8 resolved band=25 then was discarded. Guard now suppresses only when 'band === 0 && ratio < threshold' — preserves the round-to-band-0 dedup edge case while letting early-warn fire. 2. Type-required regression (Elliott): ContinuationRuntimeConfig had 'earlyWarningBand: number' (required), breaking 3 test fixtures (config.test, scheduler.test, post-compaction-delegate-dispatch.test) with TS2741. Field already optional at zod + resolver-default site; making the type optional matches. 3. Schema baseline regen (Elliott): src/config/schema.base.generated.ts needed regen to absorb the new earlyWarningBand field; preexisting models.providers.*.request.tls.insecureSkipVerify drift also absorbed in the same regen. Tests added: - checkContextPressure 'fires early-warning band below threshold when earlyWarningBand is set' (default-preservation path) - checkContextPressure 'does NOT fire below threshold when earlyWarningBand is 0' (opt-out path) All 107 affected tests pass: context-pressure (19), config (9), scheduler (12), schema.base.generated (10), post-compaction-delegate- dispatch (23), reply/context-pressure (34). Cohort cosign chain: 🩸 (root catch v5), 🌊 (default=0 catch), 🌫 (suppression-guard catch), 🌻 (type-required + baseline catch). Refs #515 --------- Co-authored-by: frond-scribe <frond-scribe@karmaterminal> Co-authored-by: Test User <test@example.com> Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> Co-authored-by: dandelion cult - cael 🩸 <cael@dandelion.cult> Co-authored-by: dandelion cult - silas 🌫 <silas.dandelion.cult@hotmail.com>

…getSessionKey absence (#463) Adds [#446] complementary trap to the [#438] mode-only trap (PR #462): - Closed-set assertion on tool.parameters.properties keys: exactly [task, delaySeconds, mode], no more, no less. Catches additions of new model-facing parameters (cross-session addressing, retry knobs, priority) without an ADR. - mode enum membership pin (normal, silent, silent-wake, post-compaction) — robust to typebox enum vs anyOf rendering. - Boolean-runtime compatibility fields (silent, silentWake, postCompaction) MUST be absent at the descriptor surface. On-disk back-compat lives in the Zod state schema (#438), not here. - targetSessionKey absence reaffirmed with load-bearing reason in the failure message (cross-session addressing is the (b)-shape lane, binary-canticle#11). Verified load-bearing: temporarily added targetSessionKey to ContinueDelegateToolSchema, the closed-set assertion failed with the expected got=[…, targetSessionKey, …] message. Restored. 9/9 passing on canonical2 cf7830f baseline. Closes #446

…pat boundary (#462) Adds structural trap test that pins: 1. Runtime objects from consumePendingDelegates and consumeStagedPostCompactionDelegates expose only `mode` — no `silent` / `silentWake` / `postCompaction` boolean runtime fields (mode-only encoding per #227). 2. continue_delegate tool descriptor advertises mode as an enum over [normal, silent, silent-wake, post-compaction] and exposes no `silent` / `silentWake` boolean parameters. 3. On-disk TaskFlow stateJson MAY still project to legacy boolean flags — back-compat positive assertion that justifies the runtime/disk encoding split. Verified load-bearing: temporarily leaking `silent: true` from flowToDelegate fails the runtime-object suite (1/9 fail), proving the trap catches the regression class #433 nearly reintroduced. 9/9 passing on canonical2 cf7830f baseline. Closes #438

silas-dandelion-cult mentioned this pull request Apr 30, 2026

[architectural-decision] Pin mode-only PendingContinuationDelegate at compat boundary #438

Open

karmafeast mentioned this pull request Apr 30, 2026

[coverage] Prove followup-runner chain state reaches disk through actual callsite #442

Closed

elliott-dandelion-cult mentioned this pull request Apr 30, 2026

test(continuation): static allowlist guard for session-keyed volatile Maps (#441) #464

Closed

karmafeast changed the base branch from main to cael/325-canonical2 April 30, 2026 06:45

silas-dandelion-cult mentioned this pull request Apr 30, 2026

test(continuation): pin request_compaction pending Set cleanup on failure + restart #465

Merged

cael-dandelion-cult changed the base branch from cael/325-canonical2 to cael/repair-step9-squash-compile April 30, 2026 06:50

cael-dandelion-cult changed the base branch from cael/repair-step9-squash-compile to cael/325-canonical2 April 30, 2026 06:50

silas-dandelion-cult mentioned this pull request Apr 30, 2026

current NTK branches - maintain me until shipped and this project closed #436

Open

elliott-dandelion-cult approved these changes May 1, 2026

View reviewed changes

silas-dandelion-cult merged commit 35c65d4 into cael/325-canonical2 May 1, 2026
4 of 12 checks passed

cael-dandelion-cult mentioned this pull request May 2, 2026

Path-B v3 cleanup waves A-E rebased onto canonical2 (cut-path for 2026.4.24) #515

Merged

elliott-dandelion-cult mentioned this pull request May 3, 2026

test(continuation): pin request_compaction registration truth-table (#445) #562

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

test(continuation): trap mode-only PendingContinuationDelegate at compat boundary#462

test(continuation): trap mode-only PendingContinuationDelegate at compat boundary#462
silas-dandelion-cult merged 1 commit intocael/325-canonical2from
silas/438-mode-only-pending-delegate-trap

silas-dandelion-cult commented Apr 30, 2026

Uh oh!

chatgpt-codex-connector Bot commented Apr 30, 2026

Uh oh!

elliott-dandelion-cult left a comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Conversation

silas-dandelion-cult commented Apr 30, 2026

What this PR adds

Why this is the trap that catches the next #433

Verified load-bearing

Test surfaces

Receipts

Refs

Uh oh!

chatgpt-codex-connector Bot commented Apr 30, 2026

💡 Codex Review

Uh oh!

elliott-dandelion-cult left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants