[renovate]

Update the compiler, formatter, analyzer, Bazel toolchain, and Renovate analyzer alias together from nightly-2026-10-18 to nightly-2026-10-19.

Clippy now enables dependency metadata and uses it when checking executable and test targets. Normal builds retain their object and native-link dependencies. This removes full Rust library generation from no-link checks where metadata is available, while retaining test-body checking and -D warnings.

The detached warm-up regression accepts typed retryable repository-discovery deferral while still requiring route publication and exactly one project open. The filesystem regression observes member and unrelated durability barriers directly, verifies readback, and checks missing-member refusal. The graph-listing ownership regression now deliberately contends the registry mutex and drives the listing and subsequent lock request with real task wakeups. It still requires no retained server lease, registry access while graph reads are blocked, and successful completion after the graph lock is released.

The current head b22521973b1a0a074a95b52de6a3e3f1e92f80e1 preserves peer integration and the shared Linux/macOS fleet-routing changes. No timeout increase or nightly rollback was made.

Validation:

• On exact source head 7487bcb222a4c5c7b257ace9ccf2a774c8c1ead0 (tree bf6fec0b9cea55c9f4edff2b520918980298e255), the full shipped release CLI build with --config=release --config=ci passed: invocation 4842725d, 937.387 seconds, 39 remote actions and 3,633 cache hits.
• The exact daemon::tests::ownership::mounted_graph_listing_takes_no_server_lease_while_it_reads test passed on that same head: 1 passed, 0 failed, 1.75 seconds, invocation f45a2468. The anti-vacuity helper confirmed execution. Teardown emitted checkpoint-scope and automation ENOENT events, so this establishes graph-listing behavior rather than clean teardown.
• Both runs forced Rustc/TestRunner remote, disabled local fallback, used priority 10, and downloaded minimal outputs. Current head b22521973b differs from this tested head only in eight .github files covering fleet routing, Tailscale network-environment preservation, and release-shell cleanup; Rust source, toolchain, manifests, and lockfiles are unchanged. Current-head hosted CI must validate those workflow changes.
• Earlier-head evidence remains historical: 23 daemon ownership tests passed; focused Clippy checked root unit tests, code-index-runtime unit tests, and filesystem integration with test-body linting under -D warnings; all six filesystem tests passed. These runs do not replace current-head CI.

Timeout investigation:

The historical asupersync action on Omarchy hit the unchanged 1,200-second limit. A fresh release compile on healthy ubuntu-main passed in 218.434 seconds, with peak resident memory 11,641,848 KiB (about 11.1 GiB), no swap, and no measured memory pressure: invocation f917a495. A single action exceeds Omarchy's 10 GiB worker cap; this supports memory contention as the timeout cause, with no compiler hang reproduced. Nightly-2026-10-19 and the timeout are preserved. No operator configuration, drain, or authentication changes were made after the user cutoff.

Full current-head CI, including actual Linux execution, its aggregate result, and the shipped CLI, remains pending before merge.

[changeset-bot]

⚠️ No Changeset found

Latest commit: b225219

Merging this PR will not cause a version bump for any packages. If these changes should not result in a new version, you're good to go. If these changes should result in a version bump, you need to add a changeset.

Click here to learn what changesets are, and how to add one.

Click here if you're a maintainer who wants to add a changeset to this PR

[renovate]

Edited/Blocked Notification

Renovate will not automatically rebase this PR, because it does not recognize the last commit author and assumes somebody else may have edited the PR.

You can manually request rebase by checking the rebase/retry box above.

⚠️ Warning: custom changes will be lost.

[ScriptedAlchemy]

Fixed the concrete memory-admission failures from native fleet job 119048856153. Its 18 code-index failures all tried to reserve a synthetic watermark band in addition to the actual text-build bytes. Applied the same reviewed canonical accounting fix as #4379: reserve the real allocation below the existing admission ceiling, with one process ledger and no budget increase. The five affected parent files were identical, so this was a clean application. Added/existing regressions cover overlapping reservations and exact cgroup headroom; pinned rustfmt, diff checks, and commitlint pass. Native CI is pending.

The session-retention-write-cost and temporal-store unit targets also timed out in that run, with no assertion error and continued progress in the logs. Their session source is unchanged from master and both pass in #4388's completed native run. I left the tests/time limits unchanged and will use the next CI result to distinguish a repeat defect from runner load.

[ScriptedAlchemy]

Pushed the remaining memory-admission convergence fix, using the reviewed implementation from #4393. Text builders now size and charge under the same canonical ledger lock, adapt to a smaller supported reservation, and keep actual admission/minimum validation inside the existing single shed-and-trim retry. Final refusal waits for the minimum working set. Removed the unused reserve_below API; no memory ceiling or retry budget was increased.

Both production files match #4393 exactly. Existing cgroup, artifact floor/refusal, charge release and reader transfer tests are retained, and the sized-admission/overlap regressions exercise the converged path. Imports are now at module scope. Pinned formatting, diff checks and commitlint pass; native behavior remains for CI. Unrelated Rust toolchain updates are preserved.