Skip to content

canary 10.0.7 - #1699

Merged
branarakic merged 104 commits into
testnet-canaryfrom
main
Jul 14, 2026
Merged

canary 10.0.7#1699
branarakic merged 104 commits into
testnet-canaryfrom
main

Conversation

@branarakic

Copy link
Copy Markdown
Contributor

Summary

Changes

Test Plan

  • Tests pass locally (pnpm test)
  • Build succeeds (pnpm build)
  • Manually tested with dkg start (if applicable)

Related Issues

Branimir Rakic and others added 30 commits July 13, 2026 18:53
…ioning

fix(cli): provision multi-arch loopback-only Blazegraph
…on-probes

fix(core): bound network isolation log storms
…tion

fix(cli): bound daemon log rotation memory
Branimir Rakic and others added 27 commits July 14, 2026 15:18
…tale event base SHA

The delta planner diffed github.event.pull_request.base.sha against the
synthetic merge commit. GitHub recomputes the merge candidate against the
CURRENT base tip, but the event payload's base.sha is a stale snapshot, so
the computed diff also contained every unrelated change merged to base since
that snapshot. Observed on PR #1690 minutes after #1687 landed: the planner
saw #1687's own cross-workspace files, reported 'Cross-cutting PR (5
production workspaces)', and routed an ordinary two-workspace PR to full CI.
On an active repo this defeats delta routing for most open PRs (fail-safe
direction only: over-testing, never under-testing).

The merge candidate's first parent IS the base tip GitHub just merged
against, and fetch-depth 2 already includes it, so derive BASE_SHA from
MERGE_SHA^1 in both planner jobs and drop the now-unneeded stale-base fetch
fallback. The planner test now pins the first-parent derivation and forbids
any reference to the stale event field in both workflows.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
…tions

Separate Context Graph discovery from edge subscriptions
…rship-harness

# Conflicts:
#	packages/agent/test/e2e-join.test.ts
…harness

fix(sync): recover private CG membership bootstrap
fix(ci): delta planner diffed against a stale base SHA, routing ordinary PRs to full CI
…en-enrollment

# Conflicts:
#	packages/agent/src/dkg-agent-join.ts
#	packages/agent/src/dkg-agent-lifecycle.ts
#	packages/agent/src/sync/responder/graph-plan.ts
#	packages/agent/test/sync-responder-oversized-fallback.test.ts
#	packages/agent/vitest.unit.config.ts
#	packages/node-ui/src/index.ts
# Conflicts:
#	packages/agent/src/dkg-agent-lifecycle.ts
#	packages/cli/src/catchup-runner-worker-impl.ts
#	packages/cli/src/catchup-runner.ts
#	packages/cli/src/daemon/routes/context-graph.ts
#	packages/cli/test/catchup-runner-worker-impl.test.ts
…ctor

The #1692 merge routed catch-up job status through
`classifyContextGraphCatchupReadiness()`. The deferral override was left
downstream of that classifier and gated on `status === "done"`, so a round
stopped by local scheduler pressure never reached it: the classifier reported
`failed`/`unreachable` first, and the readiness inspection it performs also
calls back into the agent for a round that has no readiness to inspect.

Short-circuit instead, matching the pre-merge semantics: a deferred round skips
classification entirely, never finalizes the subscription, and reports the
retryable `deferred` status. A remote denial still takes precedence — waiting on
local capacity will not clear it — and a denied-yet-deferred round still cannot
settle as a successful `done`.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
…pshot-priority

feat(sync): add configurable snapshot limits and CG priority
…ollment

feat: add bounded context graph open enrollment
…ovenance

fix(sync): invalidate stale private CG readiness proof
@branarakic
branarakic merged commit 05793cb into testnet-canary Jul 14, 2026
56 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants