This log was last generated on Wed, 30 Sep 2026 00:18:08 GMT and should not be manually modified.
Wed, 30 Sep 2026 00:18:08 GMT
- Validate a request's Rush environment before planning a process restart, so an invalid value fails the request with the native message and keeps the current daemon running.
Mon, 28 Sep 2026 20:08:26 GMT
- Drain explicit pre-execution restart results for accepted requests before closing a retiring host.
- Validate daemon startup configuration and forward idleTimeoutSeconds for the WS3 host lifecycle integration.
- Add gated native graph metadata, dependency-safe scope and invalidation controls, manual-mode pause/resume, and bounded lease-free observation using existing request and extension-event contracts.
- Add opt-in idle shutdown that protects pending requests through output drain and cleanup.
- Support protocol-negotiated acknowledged shutdown and report live process identity and resident memory in status probes.
- Bind positively identified native build/rebuild requests to a production all-project warm engine, preserve exact native selections and cache behavior, refresh inputs between requests, and reject incompatible command shapes or environments before execution.
- Add an explicitly tagged native Rushx resolver and injectable composite resolver with request-owned lifecycle children, canonical cwd/environment overrides, raw output, stdin EOF, and early pipe-close handling.
- Admit request input when its sink attaches, acknowledge drained writes, and forward EOF to owned child stdin without blocking cancellation.
- Select verified bundled or cached daemon installations and prepare exact-engine published releases with native installation APIs. Attest the actual engine and protocol in isolation before launching selected default daemon APIs.
- Make measured telemetry retention reachable through explicit production Node IPC operations. Report raw ranking inputs and exclude only proven native resource-free NoOp nodes from child-memory scoring.
- Reuse unchanged engines, reload changed configuration or command shape in-process under generation admission, and restart process-bound state through existing core ownership/startup contracts. Add isolated native install/update workers with result-before-restart ordering and no post-execution replay.
- Expose composable workspace lifecycle capabilities, preserve resolver decorators across generation reloads, dispose generation-owned resolvers, and pass explicit rushx invocations through without phased or graph interception.
- Apply daemon.watch to persistent project file observation with safe idle teardown/re-attachment, preserved runners/results and truthful cleanup accounting. Keep root/config guards and explicit request reconciliation active; never schedule autonomous scripts.
- Stop restarting the daemon when a request differs only in volatile per-shell environment variables such as PWD, OLDPWD, SHLVL, TERM or WSL_INTEROP; the restart fingerprint and the production resolver's environment check now share rush-lib's normalization.
- Fix an early coalesced phased result writing its summary before the failed operation's output summarizer was closed.
- Queue a request whose environment requires a process restart until every queued or in-flight request that matches the running process has drained, so concurrent clients with different environments no longer preempt each other.
- Report warm input-snapshot failures (with the engine diagnostics) to the failing request instead of replaying them into the next request, abort requests interrupted by a daemon shutdown with a typed DaemonShutdownError that names its initiator, and report running requests in the shutdown acknowledgement.
- Do not let the default queue timeout fail a late compatible shared build that is waiting for a running build; explicit wait limits still apply, and timeout errors say what the request waited for.
- On Linux, verify that a rushx or global command's process group has exited by reading /proc instead of running
ps --sid, so cleanup no longer fails (and retires the workspace session) on images without procpspsor with busyboxps. - Fix warmSetMaxProjects evicting the retained results of resource-free projects, which made every no-op build of a workspace with more than 20 projects re-restore the excess projects from the build cache. The cap now only counts and releases projects that hold a live runner or watcher, and ties keep the explicitly requested target over its dependencies.
- Apply
--verboseand--parallelismto the warm graph per iteration instead of reloading the graph, and only coalesce requests that share these settings. - Refresh prepared graph work under the native execution lease before resuming, preventing stale snapshots and native-command overlap.
- Wait for selected installer closure and strengthen launcher and native-mutation fixture teardown, retaining startup diagnostics and keeping fixture socket paths short.
- Await bounded Linux child-group quiescence after termination and stream closure before completing global requests, preserving immediate descendant-stop assertions and surfacing cleanup inspection failures.
- On Linux, observe watched projects with one non-recursive watch per directory created by an asynchronous walk instead of Node's recursive fs.watch emulation (one inotify watch per file, synchronous tree walk). Skip node_modules, .git, .rush/temp and declared output folders, follow directories created or removed later, tolerate directories that vanish during registration, and report inotify watch-limit exhaustion (ENOSPC) explicitly.
- Load workspace sessions and selected installations using native canonical paths, including Windows short-name aliases.
- Re-run or restore a warm operation when its declared output folders were deleted or changed outside the daemon, instead of reporting it as up to date.
- Expose the lifecycle-owned last reload tier through a shared live pong and host-status projection, without triggering initialization or deriving a tier from generation changes.
- Use ordinal ordering for request and operation identities, remove unused environment-name enumeration, and document the same-user lifecycle execution boundary.
- Terminate running operations when the last live client cancels a phased request, and answer a cancelling client immediately when other clients still need the shared work.
- Publish a coalesced phased request's result as soon as every operation in its own selection has completed and its output has drained, instead of holding it until the shared iteration finishes the other clients' larger selections.
- Keep endpoint ownership until resolver and workspace cleanup succeed, preventing successor startup against still-resident engine state.
- Acquire one native execution lease per coalesced batch, retain it through reconciliation and output/runner cleanup, and release it before results so native commands can run while the daemon is idle. Permit explicit retries after initialization contention and detect changed rig/inherited configuration before execution.
- Print the native operation summary tables and the "rush ()" line for each phased request, including warm no-op builds.
- Add a generation-owned warm-set attachment with runtime idle, project-limit, sampled memory-budget and telemetry-ranking policies; exclusive scheduler/native leases, awaited runner and scoped watcher eviction, truthful pressure diagnostics, and real native graph integration coverage.
- Preserve Windows Rushx cwd, lifecycle environment and pnpm-sync path spelling while retaining physical workspace identity, confinement and admission-time alias checks.
- Complete automatic warm-controller generation ownership, fencing cold/late attachment and awaiting pending initialization and maintenance before reload locks. Preserve ownership failures, start cold sessions with scoped watchers, expose non-initializing generation and truthful warm accounting in pong/status, and stop certifying evicted graph results from historical observer state.
- Exercise real warm Node IPC children directly on Windows, canonicalize fixture workspace paths for watcher expectations, preserve native active/protected lifetime coverage without timeout increases, and surface complete operation failures in test diagnostics.
- Retain failed request-resource ownership across result delivery, queued work, reload, shutdown and mutation restart. Keep the failed listener alive to prevent competing startup over unjoined children, while preserving ordinary failed-command restarts and successfully cleaned output failures.
- Apply request-local Rushx TTY color and width after native lifecycle environment preparation while preserving explicit non-TTY environments and native path spelling.
Tue, 22 Sep 2026 17:35:41 GMT
Version update only
Mon, 14 Sep 2026 22:42:32 GMT
- Forward reporter operation completion and iteration identity through phased request event multiplexing.
Sat, 05 Sep 2026 00:15:08 GMT
- Add authoritative Rush-compatible command result policy and ordered exact-once final result delivery for phased and global requests.
- Add an opt-in isolated execution context for caller-resolved global commands.
- Route request-scoped interactive input and signal PTY-only in-process fallback.
- Classify parsed built-in commands and admit phased and global requests through bounded workspace and graph scheduling.
- Add an opt-in phased request router that validates a caller-resolved selection, reconciles warm invalidations, runs one real graph iteration, scopes ordered streams and events to the client, and safely aborts on cancellation or disconnect.
- Merge compatible shared-build requests into one warm operation-graph iteration.
- Add an opt-in all-project engine component factory with explicit phase/plugin shape, retained invalidation reconciliation, fail-closed graph recreation boundaries, and a deterministic engine shutdown contract.
- Wire validated request lifecycles through shared warm host sessions and typed request resolvers.
Fri, 21 Aug 2026 15:16:34 GMT
Version update only
Fri, 21 Aug 2026 00:15:58 GMT
- Add a reusable warm workspace session with stable Rush configuration metadata, retained headless invalidations, and deterministic host lifecycle integration.
Thu, 20 Aug 2026 00:16:38 GMT
- Add the rushd executable and workspace-keyed daemon host bootstrap with handshake, liveness, readiness, and clean shutdown lifecycle.
- Compile package output for ES2022.
Tue, 18 Aug 2026 00:18:33 GMT
- Add the initial daemon request scheduler.