Use MikroORM's official Node SQLite dialect with bounded database isolation, preserve legacy schema and data for rollback, and require Node 22.17 or newer. Update Debian runtimes and retain ARMv6 musl compatibility without a SQLite addon.
Complete short archive writes, stop orphaned SQL workers after owner death, preserve buffered startup requests, and prestart the archived backend on every platform. Cover real package installation, migration, failure recovery, and legacy database compatibility.
Keep authentication storage failures distinct from expired sessions with a pinned Keyv runtime, propagated cache errors, and retryable HTTP 503 responses. Preserve genuine 401 failures and verify legacy cache bytes, initialization safety, and session recovery.
Package production dependencies per architecture for online and offline upgrades.
Validate dependencies and isolated backend startup before readiness, repair once
when permitted, and preserve service/data through failed preparation or rollback.
Publish release bundles with master images and expose an opt-in develop build
sharing the development index across Alpine and Debian images.
Reject invalid extra_schedules with HTTP 400 before create or update reaches the task service. Preserve omitted, null and empty extras, supported cron syntax, and additional metadata.
Localize missing rules, invalid string and array types, and non-object extra entries through the existing per-request translation middleware. Cover POST/PUT rejection without mutations, dynamic English/Chinese switching, and stable error paths.
Validation: 9 related regression tests passed on current develop; backend build passed. The API validation fix also passed full regression and isolated develop/debian-dev acceptance.
Refs #3099
Preserve supported expression fields instead of stringifying them, and probe individual fields only when whole-expression validation fails. Convert unsupported syntax and weekday ranges ending in 7 while preserving legacy date/weekday rules and valid OR branches.
Add execution regressions for issue #3099 and 64 calendar boundary cases in UTC and Asia/Shanghai. Ensure timezone child tests actually run under the Node test runner.
Validation: 745 tests passed, 8 skipped; final targeted regressions passed; backend compilation passed. Differential calendar checks matched 142,360 execution timestamps.
Fixes#3099
Download updates asynchronously with persistent progress and validate matching source and build artifacts before an explicit restart. Keep upgrade workers alive across service restarts, verify the new runtime, and recover the browser with a one-time success notice.
Publish immutable archive sizes and checksums for accurate download progress. Align the update dialog, hide unavailable actions during transfers, and show the sidebar notice only for active, ready or failed jobs. Constrain long account names with ellipsis while preserving avatar and collapse controls.
Translate UI, scheduler, validation, preload and maintenance messages dynamically, and propagate locale changes between workers. Harden update checks, downloads and terminal states, verify Debian pnpm, and ignore generated test artifacts.
Preserve literal paths and arguments, use portable command options, validate startup configuration, and retain scripts when downloads fail.
Ensure task watchdogs and process groups are cleaned up across Bash versions, with compatibility and process lifecycle regression coverage.
Map INI to the CodeMirror properties language and fall back to plain text for unsupported modes. Reap the timeout watchdog immediately under stderr redirection to prevent Bash job notifications during cleanup.
Refs #3095, #3096
Use native Gitee Pull mirrors for code and static artifacts, and verify branch and tag SHAs alongside image builds. Trigger stale mirrors through small GitHub REST ref updates, retry transient errors with backoff, and keep polling when a notification fails.
Synchronize scheduler concurrency across API and gRPC processes and restore persisted limits during recovery.
Serialize subscription mutations and preserve shared storage while cleaning Shell and CLI namespaces, historical references, and filesystem-specific naming candidates.
Handle CLI text responses correctly and include script history in backups. Add regression coverage for concurrency, cleanup ownership, long URLs, Unicode paths, and history exports.
whyour