updates:v7_1_10

v7.1.10 — update recovery and shared-history reliability

Published installer and signed pilot successor. Publication, queueing, activation and historical-data acknowledgments are separate observations.

Qualified controller release. Publication, deployment and historical-data acceptance are separate observations. See the final release receipt and published checksums for the immutable artifact hashes.

Controller update reliability

Controller updates retain their natural all-configured-rooms-empty admission, private gate, late-join cancellation and native ownership checks. The controller now records its stage before lifecycle operations and retains a safe failure code instead of discarding every exception behind a generic hold.

Two identified local cleanup conditions have bounded waits: a dedicated service process still exiting, or the existing owner lock still held. Every observation reruns the real process and stopped-ownership guards. Unsafe lock permissions, unreadable process evidence and unknown room ownership are not transient waits. The updater never kills a process to make these checks pass.

Newly admitted transactions can resume through the existing protected recovery workflow, with fresh registration, artifact, configuration and ownership checks. Automatic continuation has a durable three-attempt budget and a 30-minute eligibility window from commit admission. A healthy started successor is observed without starting it again. A recorded startup attempt with no currently verified running successor remains held for review; no uncertain room creation is automatically replayed. Legacy held jobs acquire no automatic authority. A still-running predecessor keeps its rooms while recovery restores saved reporting-service intent and returns the update to fresh empty-room admission. An owned manual timer is rearmed; sharing opt-outs are retained.

The original October 17 update failure discarded its underlying exception. Cleanup timing remains a hypothesis for that incident, not a proven cause.

Activity and historical statistics

Internal ledger identities that cannot identify an osu! player are excluded from wire events without deleting or rewriting native records. One such privacy row previously blocked the entire owner's capture batch, including historical backfill. Filtering applies to privacy, presence, joins/leaves, matches, achievements and historical baselines.

Real-player privacy preferences remain authoritative. Malformed preferences for a valid player hold ordinary uploads for review, including an older pending public preference. Correcting the preference requeues it safely. Hidden choices, sharing opt-outs, private-room exclusions, durable cursors, receipts and idempotent historical source snapshots remain in place.

Support exposes bounded exclusion counts without the excluded identities, privacy review needs and stale reporting observations. A locally queued history scan is still not proof of hub acceptance or availability in another host's .stats. Historical import continues to require an explicit reviewed source manifest.

Operations and preservation

Local status distinguishes waiting for an empty window, active update recovery and a held update that left services unexpectedly offline. The operations collector integration distinguishes stale or blocked reporting from current reporting, and respects the actual configured room topology.

Existing gameplay, commands, map-selection tuning, catalogs, storage contracts, OAuth, Discord, enrollment, website integration and operator policies are retained. Dependencies and native gameplay/storage code are unchanged. This is a controller release; it requires each host's configured rooms to empty together. A signed offer, local queue, activation and confirmed history delivery are separate observations.

Qualification

The frozen artifact passed 289 affected tests, five operations-report checks, six isolated root/service-UID cleanup cases and an exact protected upgrade from v7.1.9 preserving settings, credentials, opt-outs, catalogs, ledger and ownership. Full-catalog native lifecycle scenarios completed for three-room owner and two-room community configurations: occupied/late-join preservation, actual shutdown and delayed child cleanup, registration upgrade, two deliberate process interruptions, restart/readiness and repeated recovery without duplicate starts.

Both lifecycle runs reached the final summary write, which exposed a fixture lookup error. That reporting defect was corrected. Original failures were kept; independent read-only verification of the native jobs, interruption receipts, configuration, clean ownership and reaped processes reconciled the completed results without replaying room operations. Qualification records this explicitly.

Complete prepared catalogs contain 105,502 Standard, 19,334 Mania 4K and 4,347 Mania 7K maps plus their companions. Unchanged long gameplay/style fixtures and whole-repository discovery were not repeated. No new actual upstream room or gameplay test was performed during development; donor providers were invented and network-isolated. Live historical-data delivery still requires post-activation capture, upload acknowledgments and consumer/website observations.

Installer and checksums · Hosting and recovery · Release history

updates/v7_1_10.txt · Last modified: by wikiadmin

Donate Powered by PHP Valid HTML5 Valid CSS Driven by DokuWiki