fix: preserve post-compaction delegate arm age#488
Merged
ronan-dandelion-cult merged 1 commit intocael/325-canonical2from May 1, 2026
Merged
Conversation
63c2fca to
c657fa9
Compare
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
c657fa9 to
0cd516a
Compare
silas-dandelion-cult
added a commit
that referenced
this pull request
May 1, 2026
…tion-registration to absorb cold module-load cost
The first test in src/agents/openclaw-tools.continuation-registration.test.ts
("registers no continuation tools when continuation.enabled is unset") pays
the cold module-load cost for createOpenClawTools and its transitive imports
(compaction-attribution, pi-embedded-*, plugins/tools, config/config) under
400+ concurrent test files in the agent project.
Quiet-box first-test duration: ~95s. CI noise pushes it past vitest's 120s
per-test default, producing a flaky timeout that has now been observed across
multiple unrelated PRs:
- #485 head (compaction-attribution scope) — first-test timeout
- #488 (downstream of #485 hypothesis) — first-test timeout
- #468 head (does NOT touch this file) — same first-test, same file, timeout
Test file content is byte-identical between base cael/325-canonical2 and #485
head; the timeout is not a regression introduced by any of those PRs. Tests
2-7 in this file reuse the warm cache (~360ms each) and are unaffected.
Cure: per-test timeout bump to 240s on the first test only, with a comment
documenting the cold-start mechanism so future readers know why this single
test has a non-default timeout.
Standalone fix, deliberately not folded into #485 to keep its compaction-
attribution scope clean. Unblocks #485, #488, #468, and any future PR that
randomly trips the same flake.
Verified by silas-pr485-fixup-v2 subagent (2026-05-01 07:29 UTC):
- Local on base a3dcc2a: first test 95023ms, passed (25s margin to 120s)
- CI on #485 head 9f25f91: first test >120000ms, timed out
- CI on #468 run 25169814732: first test >120000ms, timed out (same file)
ronan-dandelion-cult
pushed a commit
that referenced
this pull request
May 1, 2026
…tion-registration to absorb cold module-load cost (#498) The first test in src/agents/openclaw-tools.continuation-registration.test.ts ("registers no continuation tools when continuation.enabled is unset") pays the cold module-load cost for createOpenClawTools and its transitive imports (compaction-attribution, pi-embedded-*, plugins/tools, config/config) under 400+ concurrent test files in the agent project. Quiet-box first-test duration: ~95s. CI noise pushes it past vitest's 120s per-test default, producing a flaky timeout that has now been observed across multiple unrelated PRs: - #485 head (compaction-attribution scope) — first-test timeout - #488 (downstream of #485 hypothesis) — first-test timeout - #468 head (does NOT touch this file) — same first-test, same file, timeout Test file content is byte-identical between base cael/325-canonical2 and #485 head; the timeout is not a regression introduced by any of those PRs. Tests 2-7 in this file reuse the warm cache (~360ms each) and are unaffected. Cure: per-test timeout bump to 240s on the first test only, with a comment documenting the cold-start mechanism so future readers know why this single test has a non-default timeout. Standalone fix, deliberately not folded into #485 to keep its compaction- attribution scope clean. Unblocks #485, #488, #468, and any future PR that randomly trips the same flake. Verified by silas-pr485-fixup-v2 subagent (2026-05-01 07:29 UTC): - Local on base a3dcc2a: first test 95023ms, passed (25s margin to 120s) - CI on #485 head 9f25f91: first test >120000ms, timed out - CI on #468 run 25169814732: first test >120000ms, timed out (same file)
1e21522
into
cael/325-canonical2
92 of 94 checks passed
silas-dandelion-cult
approved these changes
May 1, 2026
silas-dandelion-cult
left a comment
There was a problem hiding this comment.
🌫 — diff-clean approve. firstArmedAt field is purely additive (optional everywhere, falls back to stagedAt/createdAt/Date.now() when absent), schema-compatible with existing persisted state. Round-trip preserved across TaskFlow storage, post-compaction substrate test, and runner-count consume path. Test coverage includes the lossy-round-trip scenario (stagedAt: 20_000, firstArmedAt: 10_000 → consumed with firstArmedAt: 10_000). 138/15 across 10 files, all in continuation surface; no cross-cut. ✅
cael-dandelion-cult
approved these changes
May 1, 2026
cael-dandelion-cult
left a comment
There was a problem hiding this comment.
byte-walked diff — narrow, coherent: firstArmedAt threading through SessionPostCompactionDelegate / StagedPostCompactionDelegate / PendingContinuationDelegate / PendingDelegateState (zod schema), with TTL drop using stable first-armed age (requeued delegates don't reset their clock). Tests cover all three layers (canonical store, post-compaction wrapper, dispatch). POST_COMPACTION_DELEGATE_TTL_MS = 7d matches gateway max-delay shape. now injected as dep for testability. APPROVE.
Open
5 tasks
karmafeast
pushed a commit
that referenced
this pull request
May 1, 2026
…tion-registration to absorb cold module-load cost (#498) The first test in src/agents/openclaw-tools.continuation-registration.test.ts ("registers no continuation tools when continuation.enabled is unset") pays the cold module-load cost for createOpenClawTools and its transitive imports (compaction-attribution, pi-embedded-*, plugins/tools, config/config) under 400+ concurrent test files in the agent project. Quiet-box first-test duration: ~95s. CI noise pushes it past vitest's 120s per-test default, producing a flaky timeout that has now been observed across multiple unrelated PRs: - #485 head (compaction-attribution scope) — first-test timeout - #488 (downstream of #485 hypothesis) — first-test timeout - #468 head (does NOT touch this file) — same first-test, same file, timeout Test file content is byte-identical between base cael/325-canonical2 and #485 head; the timeout is not a regression introduced by any of those PRs. Tests 2-7 in this file reuse the warm cache (~360ms each) and are unaffected. Cure: per-test timeout bump to 240s on the first test only, with a comment documenting the cold-start mechanism so future readers know why this single test has a non-default timeout. Standalone fix, deliberately not folded into #485 to keep its compaction- attribution scope clean. Unblocks #485, #488, #468, and any future PR that randomly trips the same flake. Verified by silas-pr485-fixup-v2 subagent (2026-05-01 07:29 UTC): - Local on base a3dcc2a: first test 95023ms, passed (25s margin to 120s) - CI on #485 head 9f25f91: first test >120000ms, timed out - CI on #468 run 25169814732: first test >120000ms, timed out (same file)
karmafeast
added a commit
that referenced
this pull request
May 1, 2026
Co-authored-by: Test User <test@example.com> Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
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
4 participants
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.
Summary
Fixes openclaw-bootstrap#823 by preserving stable post-compaction delegate arm age and dropping stale shards before they can be queued into the fresh session.
Root cause
The TaskFlow-backed post-compaction adapter preserved shard bodies, but the legacy compatibility wrapper converted consumed staged delegates back into
SessionPostCompactionDelegatewithcreatedAt: Date.now(). That erased the original arm time whenever a shard crossed the TaskFlow/restart/re-stage boundary, making stale 10-day-old tasks look freshly armed and preventing age/TTL filtering.Fix
firstArmedAtto post-compaction delegate state and preserve it through TaskFlow state, session-store delegates, and queued delivery payloads.createdAt/firstArmedAtfrom the original arm time instead of consume time.createdAtasfirstArmedAt.firstArmedAt, with an explicit stale-drop log containing age, TTL, and task preview.Validation
pnpm test src/auto-reply/continuation/delegate-store.test.ts src/auto-reply/continuation-delegate-store.post-compaction-substrate.test.ts src/auto-reply/reply/post-compaction-delegate-dispatch.test.ts src/infra/session-delivery-queue.storage.test.tspnpm tsgo:corepnpm tsgo:core:testpnpm lintpnpm check:changedexpands to all lanes from canonical baseline unknown.agentssurfaces and fails in existing extension typecheck baselines (extensions/codexduplicate@mariozechner/pi-agent-coreidentities andextensions/qqbotzod v3/v4 mismatch), unrelated to this delegate persistence/TTL fix.