perf: Reuse compiled workspace matchers by anthonyshew · Pull Request #13970 · vercel/turborepo

anthonyshew · 2026-09-07T23:33:28Z

Description

Layer 11 of 11 in the startup-performance stack, split from #13970. Each PR is based on the previous layer so its diff contains only this optimization. The top layer preserves the original combined source tree.

Reuse compiled workspace matchers between inference and local discovery. Retain one successful set keyed by exact ordered inclusion/exclusion vectors, including spelling and duplicates. Configuration is still reread; no discovered paths, mtimes, symlink resolutions, or containment results are cached. Clone outside the mutex, preserve caller isolation, and fall back to compilation if the lock is poisoned.

Testing Instructions

Ordinary 50/500-package fixtures were mostly neutral. On a committed stress fixture with 50 real packages, one matching workspace pattern, and 200 unmatched patterns, Linux cache-hit medians improved 25.10 → 21.26 ms (15.3%) and macOS 16.42 → 15.68 ms (4.5%). Repeated --skip-infer and large dry-run guardrails did not establish a regression. DHAT cumulative allocations fell approximately 0.9–1.1 MB, with a small live-memory increase from retaining a matcher; this is not an RSS reduction. Strace confirmed unchanged filesystem-check counts. The cache is bounded to one entry, not a fixed byte budget.

Measurement limits

Measurements used optimized binaries, randomized paired before/after runs, warm OS/local caches, and synthetic npm fixtures. CPU/heap/syscall profiling was separate from wall-time timing.
Early batches had an unborn Git HEAD; later normalization and matcher-reuse batches used committed fixtures. Results from separate batches are not cumulative.
Windows runtime performance remains unmeasured. No worker-pool sizes or task concurrency defaults are reduced.

This is the original PR retained as the top of the stack. Rejected byte-scanning, direct task-output arena, and Mach-O chained-fixup experiments remain excluded. Profiling artifacts and the original combined PR description are preserved locally.

vercel · 2026-09-07T23:33:31Z

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated
examples-basic-web	Ready	Preview, v0	Sep 8, 2026 12:46pm UTC
examples-designsystem-docs	Ready	Preview, v0	Sep 8, 2026 12:46pm UTC
examples-gatsby-web	Ready	Preview, v0	Sep 8, 2026 12:46pm UTC
examples-kitchensink-blog	Ready	Preview, v0	Sep 8, 2026 12:46pm UTC
examples-nonmonorepo	Ready	Preview, v0	Sep 8, 2026 12:46pm UTC
examples-svelte-web	Ready	Preview, v0	Sep 8, 2026 12:46pm UTC
examples-tailwind-web	Ready	Preview, v0	Sep 8, 2026 12:46pm UTC
examples-vite-web	Ready	Preview, v0	Sep 8, 2026 12:46pm UTC
turbo-site	Ready	Preview, v0	Sep 8, 2026 12:46pm UTC
turborepo-factory	Ready	Preview, v0	Sep 8, 2026 12:46pm UTC

### Description Layer 1 of 11 in the startup-performance stack, split from #13970. Each PR is based on the previous layer so its diff contains only this optimization. The top layer preserves the original combined source tree. Handle help, version, and argument-error exits before creating Rayon/Tokio worker pools. Keep parsing after shim selection and diagnostics setup, so local-version selection is unchanged. Actual task execution still initializes both pools with the same concurrency settings. ### Testing Instructions Release timing of parser-only invocations showed version startup roughly 7–11% faster in the initial investigation. This is not a general `turbo run` speedup. The regression case uses an invalid Tokio worker count to ensure parser-only exits never construct a runtime. #### Measurement limits - Measurements used optimized binaries, randomized paired before/after runs, warm OS/local caches, and synthetic npm fixtures. CPU/heap/syscall profiling was separate from wall-time timing. - Early batches had an unborn Git HEAD; later normalization and matcher-reuse batches used committed fixtures. Results from separate batches are not cumulative. - Windows runtime performance remains unmeasured. No worker-pool sizes or task concurrency defaults are reduced.

### Description Undo the ten startup-optimization layers that were unintentionally merged together in #13970 (`7aaf7cbc823e5ce5579f3bc45d19abed3282505a`). This is a mechanical rollback so those changes can be reviewed and landed individually, not a rejection of the optimizations. - Keep the parser-exit optimization from the separately merged #13971. - Keep the independent Go watch-test timeout correction included in the bundled merge. - Restore the telemetry, package-manager, environment, configuration, repository-discovery, and glob code to its pre-bundle behavior. - Carry the config diagnostic-comparison test fix back with the deferred-config-check optimization when that PR is restored. ### Recovery order 1. Merge this revert first. 2. Rebuild the optimization stack on the actual merged revert commit on `main`, accounting for squash-merge ancestry. 3. Reopen #13972 through #13980 with one optimization commit per PR, and create a replacement for the matcher-reuse layer because merged #13970 cannot be reopened. 4. Review and land those draft PRs from bottom to top. Do not retarget a higher layer to `main` while its predecessors are unmerged: its branch contains their cumulative changes. The rollback is necessarily large; each reintroduced optimization will have its own isolated review diff.

### Review scope Independent optimization targeting `main`, with exactly one feature commit. No other optimization PR is a prerequisite. Recovered after the bundled #13970 merge was reverted by #13982. The separately merged parser optimization #13971 remains on `main`. This is not part of a cumulative optimization stack. The timing figures below are historical measurements from the original incremental benchmark sequence. They have not been remeasured on these newly independent branches and are not additive.  ### Description Accept ordinary numeric package-manager versions without initializing the Unicode validation regex. All other inputs use the existing parser and diagnostics, including URLs, prerelease/build labels, and non-ASCII input. Compare the fast path with the original expression in regression coverage. ### Testing Instructions DHAT attributed approximately 1 MB of startup allocations to the package-manager validation path. Bypassing it for ordinary versions removed that work; an isolated reliable wall-time improvement was not demonstrated. Do not attribute the later combined cache-hit speedups to this layer alone. #### Measurement limits - Measurements used optimized binaries, randomized paired before/after runs, warm OS/local caches, and synthetic npm fixtures. CPU/heap/syscall profiling was separate from wall-time timing. - Early batches had an unborn Git HEAD; later normalization and matcher-reuse batches used committed fixtures. Results from separate batches are not cumulative. - Windows runtime performance remains unmeasured. No worker-pool sizes or task concurrency defaults are reduced.

### Review scope Independent optimization targeting `main`, with exactly one feature commit. No other optimization PR is a prerequisite. Recovered after the bundled #13970 merge was reverted by #13982. The separately merged parser optimization #13971 remains on `main`. This is not part of a cumulative optimization stack. The timing figures below are historical measurements from the original incremental benchmark sequence. They have not been remeasured on these newly independent branches and are not additive.  ### Description Stop ancestor discovery once a multi-package root has been selected: later candidates cannot replace it under the existing selection rules. Continue searching outward when an intermediate workspace does not include the closest standalone package. ### Testing Instructions Isolated macOS measurements showed modest gains, roughly 1% for 50-package cache hits; large and forced cases were mostly inconclusive. The combined normalization/inference batch used committed synthetic Git fixtures and showed about 1.5% faster macOS 50-package hits and 3.1% faster Linux hits. Root-selection regression coverage includes nested workspaces and excluded standalone packages. Task concurrency is unchanged. #### Measurement limits - Measurements used optimized binaries, randomized paired before/after runs, warm OS/local caches, and synthetic npm fixtures. CPU/heap/syscall profiling was separate from wall-time timing. - Early batches had an unborn Git HEAD; later normalization and matcher-reuse batches used committed fixtures. Results from separate batches are not cumulative. - Windows runtime performance remains unmeasured. No worker-pool sizes or task concurrency defaults are reduced.

### Review scope Independent optimization targeting `main`, with exactly one feature commit. No other optimization PR is a prerequisite. Recovered after the bundled #13970 merge was reverted by #13982. The separately merged parser optimization #13971 remains on `main`. This is not part of a cumulative optimization stack. The timing figures below are historical measurements from the original incremental benchmark sequence. They have not been remeasured on these newly independent branches and are not additive.  ### Description Reuse compiled workspace matchers between inference and local discovery. Retain one successful set keyed by exact ordered inclusion/exclusion vectors, including spelling and duplicates. Configuration is still reread; no discovered paths, mtimes, symlink resolutions, or containment results are cached. Clone outside the mutex, preserve caller isolation, and fall back to compilation if the lock is poisoned. ### Testing Instructions Ordinary 50/500-package fixtures were mostly neutral. On a committed stress fixture with 50 real packages, one matching workspace pattern, and 200 unmatched patterns, Linux cache-hit medians improved 25.10 → 21.26 ms (15.3%) and macOS 16.42 → 15.68 ms (4.5%). Repeated `--skip-infer` and large dry-run guardrails did not establish a regression. DHAT cumulative allocations fell approximately 0.9–1.1 MB, with a small live-memory increase from retaining a matcher; this is not an RSS reduction. Strace confirmed unchanged filesystem-check counts. The cache is bounded to one entry, not a fixed byte budget. #### Measurement limits - Measurements used optimized binaries, randomized paired before/after runs, warm OS/local caches, and synthetic npm fixtures. CPU/heap/syscall profiling was separate from wall-time timing. - Early batches had an unborn Git HEAD; later normalization and matcher-reuse batches used committed fixtures. Results from separate batches are not cumulative. - Windows runtime performance remains unmeasured. No worker-pool sizes or task concurrency defaults are reduced. Rejected byte-scanning, direct task-output arena, and Mach-O chained-fixup experiments remain excluded. Profiling artifacts and the original combined PR description are preserved locally.

### Review scope Independent optimization targeting `main`, with exactly one feature commit. No other optimization PR is a prerequisite. Recovered after the bundled #13970 merge was reverted by #13982. The separately merged parser optimization #13971 remains on `main`. This is not part of a cumulative optimization stack. The timing figures below are historical measurements from the original incremental benchmark sequence. They have not been remeasured on these newly independent branches and are not additive.  ### Description Skip the normalization regexes when globstars are already standalone and non-consecutive. None of the existing rewrites can match that shape. Preserve platform path normalization and use the original rewrite chain for ambiguous patterns, including overlapping globstars. ### Testing Instructions The isolated normalization A/B showed modest Linux improvements on 50-package fixtures, with larger cases mostly neutral. A later combined batch with bounded root inference improved Linux 50-package cache hits 3.1%, per-config hits 4.3%, and dry runs 2.4%. These are incremental batches and must not be added together. Differential coverage exhaustively compares short ASCII/Unicode patterns with the original rewrites. #### Measurement limits - Measurements used optimized binaries, randomized paired before/after runs, warm OS/local caches, and synthetic npm fixtures. CPU/heap/syscall profiling was separate from wall-time timing. - Early batches had an unborn Git HEAD; later normalization and matcher-reuse batches used committed fixtures. Results from separate batches are not cumulative. - Windows runtime performance remains unmeasured. No worker-pool sizes or task concurrency defaults are reduced.

### Review scope Independent optimization targeting `main`, with exactly one feature commit. No other optimization PR is a prerequisite. Recovered after the bundled #13970 merge was reverted by #13982. The separately merged parser optimization #13971 remains on `main`. This is not part of a cumulative optimization stack. The timing figures below are historical measurements from the original incremental benchmark sequence. They have not been remeasured on these newly independent branches and are not additive.  ### Description Memoize missing workspace configs when no updater is present, and resolve the physical repository root once per loader. Preserve updater retries and synthesized configs; do not cache parse/I/O errors. Fresh loaders rediscover changed configs and symlink targets. ### Testing Instructions Linux strace on a 50-package fixture showed missing `turbo.json`/`turbo.jsonc` opens falling from 300 to 100: each of 100 distinct paths was opened once rather than three times. Root/path caching reduced readlink probes by 947 in that batch. Timing was combined with parser/environment changes, so their 15.2%/10.8% cache-hit gains are not attributed solely to this PR. This layer intentionally retains the pre-deferred-read flow; a later layer moves schema/physical checks after successful reads. #### Measurement limits - Measurements used optimized binaries, randomized paired before/after runs, warm OS/local caches, and synthetic npm fixtures. CPU/heap/syscall profiling was separate from wall-time timing. - Early batches had an unborn Git HEAD; later normalization and matcher-reuse batches used committed fixtures. Results from separate batches are not cumulative. - Windows runtime performance remains unmeasured. No worker-pool sizes or task concurrency defaults are reduced.

### Review scope Independent optimization targeting `main`, with exactly one feature commit. No other optimization PR is a prerequisite. Recovered after the bundled #13970 merge was reverted by #13982. The separately merged parser optimization #13971 remains on `main`. This is not part of a cumulative optimization stack. The timing figures below are historical measurements from the original incremental benchmark sequence. They have not been remeasured on these newly independent branches and are not additive.  ### Description Match builtin environment names/prefixes and literal-only selection lists without compiling regexes. Match builtin passthrough variables once per environment snapshot, not per task. Preserve exclusion precedence, escaping, LF behavior, and Windows Unicode case-folding through the existing regex fallback. ### Testing Instructions DHAT identified environment-matcher compilation as startup allocation work. This layer was measured together with package-manager parsing and config lookup caching: that combined Linux batch improved 50-package cache hits 15.2% and 500-package hits 10.8%. Those are combined results, not an isolated claim for environment matching. Before/after task-hash validation was used to guard cache-key compatibility. #### Measurement limits - Measurements used optimized binaries, randomized paired before/after runs, warm OS/local caches, and synthetic npm fixtures. CPU/heap/syscall profiling was separate from wall-time timing. - Early batches had an unborn Git HEAD; later normalization and matcher-reuse batches used committed fixtures. Results from separate batches are not cumulative. - Windows runtime performance remains unmeasured. No worker-pool sizes or task concurrency defaults are reduced.

### Review scope Independent optimization targeting `main`, with exactly one feature commit. No other optimization PR is a prerequisite. Recovered after the bundled #13970 merge was reverted by #13982. The separately merged parser optimization #13971 remains on `main`. This is not part of a cumulative optimization stack. The timing figures below are historical measurements from the original incremental benchmark sequence. They have not been remeasured on these newly independent branches and are not additive.  ### Description Resolve the physical repository root once per knowledge build, and reuse each manifest resolution for containment and duplicate-identity checks. Try canonicalization directly for existing files rather than statting them first; preserve the missing-path parent fallback and symlink validation. ### Testing Instructions This was measured with the following deferred-config-check layer in a separate Linux FS A/B batch. Together they reduced 500-package readlink calls from 13,042 to 6,033 and statx calls from 6,070 to 4,566; cache hits improved 6.0% and dry runs 7.2%. These are paired synthetic-fixture batch results, not isolated per-layer or universal gains. Coverage includes symlink retargeting and aliases of the same manifest. #### Measurement limits - Measurements used optimized binaries, randomized paired before/after runs, warm OS/local caches, and synthetic npm fixtures. CPU/heap/syscall profiling was separate from wall-time timing. - Early batches had an unborn Git HEAD; later normalization and matcher-reuse batches used committed fixtures. Results from separate batches are not cumulative. - Windows runtime performance remains unmeasured. No worker-pool sizes or task concurrency defaults are reduced. <sub>CLOSES TURBO-6045</sub>

### Review scope Independent optimization targeting `main`, with exactly one feature commit. No other optimization PR is a prerequisite. Recovered after the bundled #13970 merge was reverted by #13982. The separately merged parser optimization #13971 remains on `main`. This is not part of a cumulative optimization stack. The timing figures below are historical measurements from the original incremental benchmark sequence. They have not been remeasured on these newly independent branches and are not additive.  ### Description Make explicitly disabled telemetry inert. Environment opt-out skips config loading; disabled telemetry creates no worker and avoids event UUID generation, formatting, hashing, and shutdown timers. Preserve enabled behavior and distinguish opt-out from missing initialization. ### Testing Instructions The startup investigation traced work performed even when telemetry was disabled. End-to-end run measurements did not establish a consistent standalone speedup for this layer; this removes avoidable opt-out work rather than claiming the combined benchmark gains. Enabled-mode measurements used a closed local proxy to exclude external network variability. #### Measurement limits - Measurements used optimized binaries, randomized paired before/after runs, warm OS/local caches, and synthetic npm fixtures. CPU/heap/syscall profiling was separate from wall-time timing. - Early batches had an unborn Git HEAD; later normalization and matcher-reuse batches used committed fixtures. Results from separate batches are not cumulative. - Windows runtime performance remains unmeasured. No worker-pool sizes or task concurrency defaults are reduced.

### Review scope Depends on #13975: this change uses its per-loader `physical_repo_root` cache and extends its missing-config coverage. This PR targets `shew/startup-05-config-lookups`, so its diff contains only deferred filesystem checks. After #13975 lands, rebase this feature commit onto updated `main` before retargeting it. No other optimization PR is a prerequisite. Recovered after the bundled #13970 merge was reverted by #13982. The separately merged parser optimization #13971 remains on `main`. This is not part of a cumulative optimization stack. The timing figures below are historical measurements from the original incremental benchmark sequence. They have not been remeasured on these newly independent branches and are not additive.  ### Description Resolve root-versus-package config schema only after reading config contents. Reuse explicit config-existence observations while preserving precedence and errors for conflicting configs. Avoid physical-path/schema work for absent configs without removing checks for files that are actually read. ### Testing Instructions In the Linux FS batch with the preceding repository-canonicalization layer, 500-package cache hits improved 6.0% and dry runs 7.2%. Smaller and forced-run cases were inconclusive, including a 2.7% median slowdown for the per-package-config cache-hit fixture whose paired interval spanned zero. All 888 saved timing measurements were preserved when only the unfinished strace/perf phase was resumed. Do not add these percentages to earlier batches. #### Measurement limits - Measurements used optimized binaries, randomized paired before/after runs, warm OS/local caches, and synthetic npm fixtures. CPU/heap/syscall profiling was separate from wall-time timing. - Early batches had an unborn Git HEAD; later normalization and matcher-reuse batches used committed fixtures. Results from separate batches are not cumulative. - Windows runtime performance remains unmeasured. No worker-pool sizes or task concurrency defaults are reduced.

### Review scope Independent optimization targeting `main`, with exactly one feature commit. No other optimization PR is a prerequisite. Recovered after the bundled #13970 merge was reverted by #13982. The separately merged parser optimization #13971 remains on `main`. This is not part of a cumulative optimization stack. The timing figures below are historical measurements from the original incremental benchmark sequence. They have not been remeasured on these newly independent branches and are not additive.  ### Description Compile workspace directory-glob combinators directly from expressions instead of first compiling individual regexes that `wax::any` discards. Add an owning conversion for `Any` without recompiling. Retain the original per-expression validation/error fallback. ### Testing Instructions DHAT showed another approximately 0.83 MB fewer allocations per 50-package invocation. Incremental Linux wall times were mostly neutral, with a modest roughly 2% improvement on the 500-package cache-hit fixture. Matching and task hashes were unchanged. This is not a speedup equal to the full workspace-compilation share in the CPU profile. #### Measurement limits - Measurements used optimized binaries, randomized paired before/after runs, warm OS/local caches, and synthetic npm fixtures. CPU/heap/syscall profiling was separate from wall-time timing. - Early batches had an unborn Git HEAD; later normalization and matcher-reuse batches used committed fixtures. Results from separate batches are not cumulative. - Windows runtime performance remains unmeasured. No worker-pool sizes or task concurrency defaults are reduced.

## Release v2.10.13-canary.2 > [!CAUTION] > Versioned docs aliasing FAILED. [View logs](https://github.com/vercel/turborepo/actions/runs/34359600232) ### Changes - fix: Keep Cargo tasks runnable without compiler identity (#13857) (`31645fb`) - chore: Release Turborepo 2.10.13-canary.1 (#13858) (`5ac6ea4`) - feat: Support Cargo root packages (#13856) (`e732034`) - feat(repository): Expose flat lockfile package list for metrics (#13859) (`935ee2c`) - chore: Release Turbo repository packages 0.0.1-canary.25 (#13860) (`b0d4c71`) - docs: Upgrade Geistdocs to 1.25.1 (#13861) (`81b0414`) - feat: Expand repository lockfile package metadata (#13862) (`2c84c66`) - chore: Release Turbo repository packages 0.0.1-canary.26 (#13863) (`69e394b`) - fix: Restrict Factory to review comments (#13864) (`1a34319`) - fix: Pluralize package count in run prelude (#13865) (`a175bd9`) - chore: Update with-vite example (#13866) (`0e1ff63`) - chore: Add searchable workspace model picker (#13870) (`73b6268`) - fix: Restore docs client navigation (#13872) (`2995e17`) - feat: Improve Factory workspace statuses (#13869) (`35f03fa`) - chore: Clarify Factory pull request descriptions (#13871) (`5df4cb3`) - fix: Refresh Factory workspaces from main (#13873) (`efa4ae0`) - perf: Speed up plain package listings (#13868) (`f68d214`) - fix: Detect package-level turbo.jsonc (#13875) (`e945403`) - fix: Format Slack pull request notifications (#13877) (`15c15f3`) - feat: Link workspaces to pull requests (#13878) (`ae25f97`) - chore: Upgrade repository to pnpm 12 (#13879) (`4e11bad`) - chore: Update with-vite-react example (#13880) (`54cf466`) - chore: Improve Factory chat streaming (#13881) (`eebf1b0`) - fix: Show workspace failure details (#13884) (`c7661b8`) - chore: Add selectable Factory coding harnesses (#13887) (`6b5ca16`) - fix: Gracefully handle native lockfile failures (#13886) (`463a292`) - chore: Update merged PR Slack messages (#13888) (`66ebcac`) - fix: Show native tasks in task listings (#13889) (`7e4d9d3`) - docs: Clarify turbo ls stability in Agent Skill (#13892) (`8ee758c`) - chore: Resume workspace chat on refocus (#13894) (`7026627`) - chore: Preserve HarnessAgent bridge assets (#13896) (`26f3160`) - fix: Deduplicate multi-language watch hashes (#13897) (`5ade89b`) - fix: Normalize multi-language cache environment inputs (#13898) (`cfca6f2`) - chore: Add workspace diff view (#13900) (`813d54a`) - refactor: Migrate CLI parser to usage-rs (#13890) (`712e0e4`) - fix: Update --filter help links to canonical docs URL (#13891) (`6cf116d`) - chore: Update with-vue-nuxt example (#13901) (`d10e3e0`) - docs: Document missing reference flags and fix query typo (#13902) (`1f2acc5`) - docs: Align CONTRIBUTING Node and pnpm versions (#13883) (`a3bda61`) - perf: Skip Cargo resolution for workspace-only queries (#13903) (`b928ecc`) - fix: Bound memory usage for streamed task output (#13908) (`7f85401`) - fix: Clarify generator action reporting (#13910) (`7fe0201`) - fix: Bump tar to 7.5.22 in @turbo/releaser (CVE-2026-73566) (#13911) (`aba64e9`) - fix: Correct devtools --port default in CLI help (#13905) (`11f7ee6`) - fix: Resolve lint errors (#13912) (`d06cf32`) - fix: Stabilize Rust builds on macOS (#13913) (`7d14df2`) - docs: Clean up Rust docs (#13914) (`3225cee`) - fix: Preserve task metadata for command overrides (#13915) (`54bbf8a`) - perf: Resolve Git SCM state in one process (#13882) (`7c993ef`) - fix: Only treat files literally named .gitignore as ignore files in the untracked walk (#13916) (`ac32a79`) - fix: Pass merge_base and allow_unknown_objects in signature order for task-level filter ranges (#13917) (`48f64e1`) - fix: Hash deferred inputs after dependencies (#13919) (`4cc80f7`) - test: Make Cargo hash fixtures hermetic (#13920) (`32ff2f3`) - fix: Allow release versioning with pending changes (#13921) (`5114901`) - chore: Update to Rust 1.98.0 (#13821) (`ebef942`) - chore: Upgrade Factory to Fable 5.1 (#13924) (`a57476f`) - docs: Align Python guide with Rust guide (#13927) (`f166370`) - fix: Give the package changes watcher the future flags so root inputs reach the task filter (#13926) (`e5ccf55`) - fix: Support pnpm 6 single-package lockfiles (#13928) (`a3d52a6`) - feat: Add skipPackageGraph option to @turbo/repository Workspace.find (#13929) (`99b5a00`) - fix: Use pnpm 12 in library release containers (#13930) (`0b6aaf9`) - fix: Upgrade repository N-API build tooling (#13931) (`dd6c1c6`) - fix: Use Node 20 for musl library builds (#13933) (`61b1635`) - fix: Bootstrap musl builds with sh (#13934) (`f9ac0c9`) - fix: Use static library bootstrap shell (#13935) (`f819dca`) - fix: Allow unclean tree in bump-version (#13936) (`28562e7`) - fix: Skip git commit in bump-version (#13937) (`ea4ba17`) - fix: Expect napi 3 artifact paths (#13938) (`6b5eea3`) - chore: Release Turbo repository packages 0.0.1-canary.27 (#13939) (`5c54e50`) - perf: Batch package detail queries (#13923) (`c8347b1`) - docs: Touch-ups for Python and Rust docs (#13940) (`f20b415`) - chore: Update basic example (#13922) (`b40dd07`) - fix: Avoid oxlint false positive on Workspace.find options (#13941) (`3125eb1`) - chore: Update kitchen-sink example (#13945) (`fd8dc59`) - fix: Hash native workspace task inputs (#13947) (`17d0940`) - chore: Update with-angular example (#13951) (`3770d4c`) - fix: Recompute eager hashing flag after prepending global inputs (#13949) (`9cea33b`) - refactor: Generalize daemon repository discovery (#13946) (`350bd41`) - feat: Make daemon discovery toolchain generic (#13952) (`d65c615`) - feat: Add experimental Go workspace graph foundation (#13953) (`6687fa8`) - feat: Add native Go task tables (#13954) (`dccc890`) - feat: Add Go task input and output contracts (#13955) (`8501ce7`) - test: Cover native Go execution end to end (#13956) (`48d4322`) - feat: Add Go external resolution fingerprints (#13957) (`f6d5f18`) - feat: Fingerprint Go toolchain environment (#13958) (`ec18363`) - feat: Add Go change observations (#13959) (`5256ecc`) - test: Cover Go query and summary surfaces (#13960) (`a24ea48`) - feat: Add Go-aware prune planning (#13961) (`026d7fb`) - test: Add end-to-end Go prune and cache coverage (#13962) (`a1cbfaa`) - docs: Document experimental native Go workspaces (#13963) (`7f36a42`) - fix: Harden Go workspace diagnostics (#13964) (`8a6b0f7`) - fix: Correct Go workspace resolution and task caching (#13966) (`8ef6aaf`) - chore: Add Factory thinking controls and message queue (#13967) (`77e975c`) - chore: Remove ARCHITECTURE.md docs (#13969) (`4f8e693`) - fix: Add slug to OPTIONS request (#13943) (`63797e4`) - perf: Parse CLI exits before starting worker pools (#13971) (`6888882`) - perf: Reuse compiled workspace matchers (#13970) (`7aaf7cb`) - chore: Revert bundled startup optimizations (#13982) (`c11e327`) - perf: Fast-path numeric package manager versions (#13973) (`3b28640`) - perf: Stop inference after selecting a workspace root (#13980) (`a3c5b9c`) - perf: Reuse compiled workspace matchers (#13984) (`eb80d6f`) - perf: Skip unnecessary glob normalization (#13979) (`1840e84`) - perf: Cache workspace configuration lookups (#13975) (`f59661a`) - perf: Avoid regexes for simple environment matching (#13974) (`7e5d793`) - perf: Reuse canonical paths during repository discovery (#13976) (`2ba3e6e`) - perf: Skip work for disabled telemetry (#13972) (`267ed73`) - perf: Defer filesystem checks for absent configs (#13977) (`f7db9df`) - perf: Compile workspace glob combinators directly (#13978) (`223fa9f`) - docs(skills): fix the dependencies-only filter caret position (#13965) (`73ac1fa`) - chore: Update with-biome example (#13983) (`bd331cf`) - fix: Normalize bare directory outputs when filtering dependencyOutputs hashes (#13932) (`555f2bc`) - chore: Scope panic lint allows in hash crates to tests and generated code (#13810) (`e901550`) - perf: Use binary search for resolution states (#13986) (`6a7f1c6`) - perf: Use dense membership for task graph pruning (#13987) (`4ffbaa9`) - perf: Throttle TUI tick cadence (#13988) (`b74be22`) - perf: Fast-path package-only filter selectors with --only (#13990) (`8059b91`) - perf: Avoid materializing npm's legacy dependency tree (#13993) (`acd35c8`) - perf: Stream custom repository downloads to disk (#13994) (`a6b88ee`) - perf: Copy shared file dependencies once per destination during prune (#13995) (`efbe558`) - perf: Index npm workspace links during subgraph extraction (#13997) (`996a0bb`) - perf: Cache Turbo config reads across ESLint rule invocations (#13998) (`c02f97e`) - perf: Scope untracked-file discovery to filtered runs (#13991) (`ead0e19`) - perf: Limit turbo.json preloading to required task scope (#13992) (`f788cd8`) - ci: Restrict release signing to main (#13985) (`2258541`) - perf: Coalesce glob-watcher invalidation snapshots (#13996) (`3fec2fe`) - perf: Reuse resolver inference during file tracing (#13999) (`596ea4c`) Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>

## Release v2.10.13 > [!CAUTION] > Versioned docs aliasing FAILED. [View logs](https://github.com/vercel/turborepo/actions/runs/34866964811) ### Changes - chore: Release Turborepo 2.10.12 (#13844) (`32748f5`) - fix: Remove unsupported remote cache environment variable (#13845) (`b4c2eed`) - fix: Copy TUI selections locally over SSH (#13847) (`03df632`) - feat: Use uv workspace metadata (#13848) (`fa1ca7d`) - feat: Support Python virtual environments (#13849) (`7f66dbd`) - fix: Scope uv lockfile affectedness (#13850) (`1e074f3`) - test: Isolate uv prune configuration (#13851) (`9f2fd33`) - fix: Explain disabled uv task caching (#13852) (`0f59d11`) - fix: Explain uv identity probe failures (#13853) (`eabe73a`) - fix: Explain uncached Cargo library builds (#13855) (`35ce2fa`) - fix: Explain disabled Cargo task caching (#13854) (`39821f6`) - fix: Keep Cargo tasks runnable without compiler identity (#13857) (`31645fb`) - chore: Release Turborepo 2.10.13-canary.1 (#13858) (`5ac6ea4`) - feat: Support Cargo root packages (#13856) (`e732034`) - feat(repository): Expose flat lockfile package list for metrics (#13859) (`935ee2c`) - chore: Release Turbo repository packages 0.0.1-canary.25 (#13860) (`b0d4c71`) - docs: Upgrade Geistdocs to 1.25.1 (#13861) (`81b0414`) - feat: Expand repository lockfile package metadata (#13862) (`2c84c66`) - chore: Release Turbo repository packages 0.0.1-canary.26 (#13863) (`69e394b`) - fix: Restrict Factory to review comments (#13864) (`1a34319`) - fix: Pluralize package count in run prelude (#13865) (`a175bd9`) - chore: Update with-vite example (#13866) (`0e1ff63`) - chore: Add searchable workspace model picker (#13870) (`73b6268`) - fix: Restore docs client navigation (#13872) (`2995e17`) - feat: Improve Factory workspace statuses (#13869) (`35f03fa`) - chore: Clarify Factory pull request descriptions (#13871) (`5df4cb3`) - fix: Refresh Factory workspaces from main (#13873) (`efa4ae0`) - perf: Speed up plain package listings (#13868) (`f68d214`) - fix: Detect package-level turbo.jsonc (#13875) (`e945403`) - fix: Format Slack pull request notifications (#13877) (`15c15f3`) - feat: Link workspaces to pull requests (#13878) (`ae25f97`) - chore: Upgrade repository to pnpm 12 (#13879) (`4e11bad`) - chore: Update with-vite-react example (#13880) (`54cf466`) - chore: Improve Factory chat streaming (#13881) (`eebf1b0`) - fix: Show workspace failure details (#13884) (`c7661b8`) - chore: Add selectable Factory coding harnesses (#13887) (`6b5ca16`) - fix: Gracefully handle native lockfile failures (#13886) (`463a292`) - chore: Update merged PR Slack messages (#13888) (`66ebcac`) - fix: Show native tasks in task listings (#13889) (`7e4d9d3`) - docs: Clarify turbo ls stability in Agent Skill (#13892) (`8ee758c`) - chore: Resume workspace chat on refocus (#13894) (`7026627`) - chore: Preserve HarnessAgent bridge assets (#13896) (`26f3160`) - fix: Deduplicate multi-language watch hashes (#13897) (`5ade89b`) - fix: Normalize multi-language cache environment inputs (#13898) (`cfca6f2`) - chore: Add workspace diff view (#13900) (`813d54a`) - refactor: Migrate CLI parser to usage-rs (#13890) (`712e0e4`) - fix: Update --filter help links to canonical docs URL (#13891) (`6cf116d`) - chore: Update with-vue-nuxt example (#13901) (`d10e3e0`) - docs: Document missing reference flags and fix query typo (#13902) (`1f2acc5`) - docs: Align CONTRIBUTING Node and pnpm versions (#13883) (`a3bda61`) - perf: Skip Cargo resolution for workspace-only queries (#13903) (`b928ecc`) - fix: Bound memory usage for streamed task output (#13908) (`7f85401`) - fix: Clarify generator action reporting (#13910) (`7fe0201`) - fix: Bump tar to 7.5.22 in @turbo/releaser (CVE-2026-73566) (#13911) (`aba64e9`) - fix: Correct devtools --port default in CLI help (#13905) (`11f7ee6`) - fix: Resolve lint errors (#13912) (`d06cf32`) - fix: Stabilize Rust builds on macOS (#13913) (`7d14df2`) - docs: Clean up Rust docs (#13914) (`3225cee`) - fix: Preserve task metadata for command overrides (#13915) (`54bbf8a`) - perf: Resolve Git SCM state in one process (#13882) (`7c993ef`) - fix: Only treat files literally named .gitignore as ignore files in the untracked walk (#13916) (`ac32a79`) - fix: Pass merge_base and allow_unknown_objects in signature order for task-level filter ranges (#13917) (`48f64e1`) - fix: Hash deferred inputs after dependencies (#13919) (`4cc80f7`) - test: Make Cargo hash fixtures hermetic (#13920) (`32ff2f3`) - fix: Allow release versioning with pending changes (#13921) (`5114901`) - chore: Update to Rust 1.98.0 (#13821) (`ebef942`) - chore: Upgrade Factory to Fable 5.1 (#13924) (`a57476f`) - docs: Align Python guide with Rust guide (#13927) (`f166370`) - fix: Give the package changes watcher the future flags so root inputs reach the task filter (#13926) (`e5ccf55`) - fix: Support pnpm 6 single-package lockfiles (#13928) (`a3d52a6`) - feat: Add skipPackageGraph option to @turbo/repository Workspace.find (#13929) (`99b5a00`) - fix: Use pnpm 12 in library release containers (#13930) (`0b6aaf9`) - fix: Upgrade repository N-API build tooling (#13931) (`dd6c1c6`) - fix: Use Node 20 for musl library builds (#13933) (`61b1635`) - fix: Bootstrap musl builds with sh (#13934) (`f9ac0c9`) - fix: Use static library bootstrap shell (#13935) (`f819dca`) - fix: Allow unclean tree in bump-version (#13936) (`28562e7`) - fix: Skip git commit in bump-version (#13937) (`ea4ba17`) - fix: Expect napi 3 artifact paths (#13938) (`6b5eea3`) - chore: Release Turbo repository packages 0.0.1-canary.27 (#13939) (`5c54e50`) - perf: Batch package detail queries (#13923) (`c8347b1`) - docs: Touch-ups for Python and Rust docs (#13940) (`f20b415`) - chore: Update basic example (#13922) (`b40dd07`) - fix: Avoid oxlint false positive on Workspace.find options (#13941) (`3125eb1`) - chore: Update kitchen-sink example (#13945) (`fd8dc59`) - fix: Hash native workspace task inputs (#13947) (`17d0940`) - chore: Update with-angular example (#13951) (`3770d4c`) - fix: Recompute eager hashing flag after prepending global inputs (#13949) (`9cea33b`) - refactor: Generalize daemon repository discovery (#13946) (`350bd41`) - feat: Make daemon discovery toolchain generic (#13952) (`d65c615`) - feat: Add experimental Go workspace graph foundation (#13953) (`6687fa8`) - feat: Add native Go task tables (#13954) (`dccc890`) - feat: Add Go task input and output contracts (#13955) (`8501ce7`) - test: Cover native Go execution end to end (#13956) (`48d4322`) - feat: Add Go external resolution fingerprints (#13957) (`f6d5f18`) - feat: Fingerprint Go toolchain environment (#13958) (`ec18363`) - feat: Add Go change observations (#13959) (`5256ecc`) - test: Cover Go query and summary surfaces (#13960) (`a24ea48`) - feat: Add Go-aware prune planning (#13961) (`026d7fb`) - test: Add end-to-end Go prune and cache coverage (#13962) (`a1cbfaa`) - docs: Document experimental native Go workspaces (#13963) (`7f36a42`) - fix: Harden Go workspace diagnostics (#13964) (`8a6b0f7`) - fix: Correct Go workspace resolution and task caching (#13966) (`8ef6aaf`) - chore: Add Factory thinking controls and message queue (#13967) (`77e975c`) - chore: Remove ARCHITECTURE.md docs (#13969) (`4f8e693`) - fix: Add slug to OPTIONS request (#13943) (`63797e4`) - perf: Parse CLI exits before starting worker pools (#13971) (`6888882`) - perf: Reuse compiled workspace matchers (#13970) (`7aaf7cb`) - chore: Revert bundled startup optimizations (#13982) (`c11e327`) - perf: Fast-path numeric package manager versions (#13973) (`3b28640`) - perf: Stop inference after selecting a workspace root (#13980) (`a3c5b9c`) - perf: Reuse compiled workspace matchers (#13984) (`eb80d6f`) - perf: Skip unnecessary glob normalization (#13979) (`1840e84`) - perf: Cache workspace configuration lookups (#13975) (`f59661a`) - perf: Avoid regexes for simple environment matching (#13974) (`7e5d793`) - perf: Reuse canonical paths during repository discovery (#13976) (`2ba3e6e`) - perf: Skip work for disabled telemetry (#13972) (`267ed73`) - perf: Defer filesystem checks for absent configs (#13977) (`f7db9df`) - perf: Compile workspace glob combinators directly (#13978) (`223fa9f`) - docs(skills): fix the dependencies-only filter caret position (#13965) (`73ac1fa`) - chore: Update with-biome example (#13983) (`bd331cf`) - fix: Normalize bare directory outputs when filtering dependencyOutputs hashes (#13932) (`555f2bc`) - chore: Scope panic lint allows in hash crates to tests and generated code (#13810) (`e901550`) - perf: Use binary search for resolution states (#13986) (`6a7f1c6`) - perf: Use dense membership for task graph pruning (#13987) (`4ffbaa9`) - perf: Throttle TUI tick cadence (#13988) (`b74be22`) - perf: Fast-path package-only filter selectors with --only (#13990) (`8059b91`) - perf: Avoid materializing npm's legacy dependency tree (#13993) (`acd35c8`) - perf: Stream custom repository downloads to disk (#13994) (`a6b88ee`) - perf: Copy shared file dependencies once per destination during prune (#13995) (`efbe558`) - perf: Index npm workspace links during subgraph extraction (#13997) (`996a0bb`) - perf: Cache Turbo config reads across ESLint rule invocations (#13998) (`c02f97e`) - perf: Scope untracked-file discovery to filtered runs (#13991) (`ead0e19`) - perf: Limit turbo.json preloading to required task scope (#13992) (`f788cd8`) - ci: Restrict release signing to main (#13985) (`2258541`) - perf: Coalesce glob-watcher invalidation snapshots (#13996) (`3fec2fe`) - perf: Reuse resolver inference during file tracing (#13999) (`596ea4c`) - chore: Release Turborepo 2.10.13-canary.2 (#14000) (`ebfc808`) - chore: Exclude turbo and @turbo/* from minimum release age (#14005) (`5897420`) - perf: Binary search boundaries comments before each import (#14003) (`6a95c81`) - perf: Fetch only npm dist tags in the binary version endpoint (#14007) (`4db9ca0`) - perf: Compute Devtools graph depths without rescanning every edge (#14001) (`ef46772`) - perf: Reuse package manager detection during create startup (#14006) (`06f5900`) - perf: Borrow workspace entries during dependency lookup (#14002) (`ba822e1`) - perf: Cache LSP task index and dedupe identities in constant time (#14008) (`77f6232`) - perf: Deduplicate recursive glob prefixes before walking (#14004) (`90fd2d6`) - perf: Batch boundaries progress updates (#14009) (`3862f3e`) - perf: Reuse archive anchor during cache creation (#14010) (`f01afa9`) - perf: Bound remote-cache transfer memory instead of retaining whole artifacts (#14011) (`470ffb9`) - perf: Build cache archives once for local and remote writes (#14012) (`f0a0d77`) - perf: Make VS Code binary discovery and LSP probes nonblocking (#14015) (`8cc0c42`) - perf: Avoid serializing ASTs for path-only file dependency queries (#14016) (`96712c9`) - perf: Share source text across boundaries diagnostics instead of copying it per error (#14014) (`5223bd6`) - perf: Parse the lockfile once when opening a native Workspace with its graph (#14017) (`23ee440`) - docs: Document a focused Cargo build for CLI-only development (#14018) (`d21277e`) - perf: Run the docs MDX generator once per quality task graph (#14019) (`f7b030b`) - perf: Keep package bundling inside a cacheable build boundary (#14020) (`f3ae66f`) - perf: Skip speculative content hashing for size-changed unnormalized tracked files (#14021) (`a8b4ac2`) - perf: Move synchronous cache archiving off Tokio runtime workers (#14013) (`22b6577`) - perf: Index package directories for changed-file ownership lookups (#14022) (`05f16eb`) - perf: Bound TUI raw-output retention independently of scrollback (#14024) (`0cb6d58`) - perf: Coalesce package rediscovery while a scan is already running (#14023) (`c6eac09`) - perf: Replace repeated deferred-hash scans with dependency-driven readiness (#14025) (`9a98e64`) - perf: Read child output in bounded chunks instead of per-line (#14026) (`f6214c8`) - perf: Share one repo index across same-package spec rehashes in watch mode (#14027) (`171a0a3`) - test: Deflake rediscovery_coalesces_during_in_flight_scan (#14029) (`d41369b`) - perf: Reuse the package graph across partial watch reruns (#14028) (`9a42fe3`) - test: Deflake Go watch CI (#14031) (`0af9b4f`) - chore: Release Turborepo 2.10.13-canary.3 (#14032) (`cfd82d2`) - docs: Group experimental languages and link the Go RFC (#14034) (`6e8f7fe`) - chore: Update with-docker example (#14035) (`25fe817`) - perf: Avoid cloning package list in turbo ls (#14036) (`f3b0fa9`) - fix: Restore leading flags for implicit run commands (#14037) (`3235ec5`) - fix: Include experimentalCI in task hashes (#14038) (`b8c7370`) - chore: Release Turborepo 2.10.13-canary.4 (#14039) (`17b358a`) - docs: Explain circular package dependencies in boundaries (#14042) (`895337e`) - docs: Document circular dependencies in boundaries (#14041) (`3821e42`) - fix: Preserve Yarn resolution declaration order (#14045) (`54b74a7`) - fix: Avoid duplicate interrupts during task shutdown (#14046) (`f9fa885`) - fix: Preserve nested Bun package keys (#14050) (`b647afc`) - chore: Update with-mcp-servers example (#14049) (`2c11c4f`) - chore: Release Turborepo 2.10.13-canary.5 (#14051) (`cc1caca`) - fix: Preserve native task dependencies across entrypoints (#14047) (`ce4f045`) - fix: Pass through LOCALAPPDATA by default (#14056) (`258a08d`) - chore: Release Turborepo 2.10.13-canary.6 (#14057) (`bea9b6c`) - docs: Fix some typos (#14058) (`6a3b4c1`) - feat: Add Remix inference and `NUXT_ENV_*` wildcard (#14060) (`2167e74`) - fix: Return resolved package names for aliases (#14061) (`83f3ee1`) - chore: Release Turbo repository packages 0.0.1-canary.28 (#14062) (`e5094c7`) - feat: Add NEXT_DEPLOYMENT_ID to Next.js framework inference (#14065) (`36383cb`) --------- Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>

## Release v2.11.0 > [!CAUTION] > Versioned docs aliasing FAILED. [View logs](https://github.com/vercel/turborepo/actions/runs/35366297249) ### Changes - feat: Add durable Factory workspaces (#13804) (`ccd79d3`) - feat: Stream Factory sandbox as terminal (#13807) (`33d8b24`) - fix: Restore Factory workspace creation (#13812) (`fd72cca`) - fix: Update Factory session network policy (#13814) (`c579fec`) - fix: Improve Factory terminal line spacing (#13815) (`a88e39b`) - fix: Install Factory publishing skill (#13817) (`7b8cb14`) - chore: Route Factory publishing through Eve (#13816) (`e05b81c`) - feat: Standardize Factory meta titles to Turborepo suffix (#13819) (`fd593b5`) - perf: Skip Factory chat verification (#13820) (`39137a6`) - fix: Restore Factory workspaces (#13823) (`0afd4b2`) - docs: Fix inconsistent Yarn command in basic example (#13825) (`3dd49d1`) - chore: Update non-monorepo example (#13808) (`d2a673f`) - perf: Batch package detail queries (#13809) (`331183e`) - chore: Update basic example (#13824) (`f06836d`) - fix: Include virtual tasks in affected query (#13805) (`89a9b78`) - fix: Add workspace approval controls (#13827) (`02dfd21`) - fix: Prevent chat SSH command overflow (#13828) (`d9eaff5`) - fix: Update Factory pull request branches (#13831) (`b670754`) - chore: Add operator chat model selector (#13833) (`bb2fcb3`) - fix: Move model selector to workspace creation (#13835) (`72805d7`) - chore: Skip redundant Factory PR approval (#13837) (`bbe5406`) - chore: Use geistdocs 1.23.1 (#13834) (`3787c06`) - chore: Handle feedback on Factory pull requests (#13836) (`a76330b`) - fix: Escape ampersands in RSS feed enclosure URLs (#13839) (`7107f26`) - chore: Add automatic issue handling (#13840) (`e1674e4`) - chore: Alert Slack for low-confidence issues (#13841) (`f153cda`) - feat: Require high confidence for issue fixes (#13842) (`c97782b`) - fix: Run pnpm directly on Windows (#13843) (`9d2b03b`) - chore: Release Turborepo 2.10.12 (#13844) (`32748f5`) - fix: Remove unsupported remote cache environment variable (#13845) (`b4c2eed`) - fix: Copy TUI selections locally over SSH (#13847) (`03df632`) - feat: Use uv workspace metadata (#13848) (`fa1ca7d`) - feat: Support Python virtual environments (#13849) (`7f66dbd`) - fix: Scope uv lockfile affectedness (#13850) (`1e074f3`) - test: Isolate uv prune configuration (#13851) (`9f2fd33`) - fix: Explain disabled uv task caching (#13852) (`0f59d11`) - fix: Explain uv identity probe failures (#13853) (`eabe73a`) - fix: Explain uncached Cargo library builds (#13855) (`35ce2fa`) - fix: Explain disabled Cargo task caching (#13854) (`39821f6`) - fix: Keep Cargo tasks runnable without compiler identity (#13857) (`31645fb`) - chore: Release Turborepo 2.10.13-canary.1 (#13858) (`5ac6ea4`) - feat: Support Cargo root packages (#13856) (`e732034`) - feat(repository): Expose flat lockfile package list for metrics (#13859) (`935ee2c`) - chore: Release Turbo repository packages 0.0.1-canary.25 (#13860) (`b0d4c71`) - docs: Upgrade Geistdocs to 1.25.1 (#13861) (`81b0414`) - feat: Expand repository lockfile package metadata (#13862) (`2c84c66`) - chore: Release Turbo repository packages 0.0.1-canary.26 (#13863) (`69e394b`) - fix: Restrict Factory to review comments (#13864) (`1a34319`) - fix: Pluralize package count in run prelude (#13865) (`a175bd9`) - chore: Update with-vite example (#13866) (`0e1ff63`) - chore: Add searchable workspace model picker (#13870) (`73b6268`) - fix: Restore docs client navigation (#13872) (`2995e17`) - feat: Improve Factory workspace statuses (#13869) (`35f03fa`) - chore: Clarify Factory pull request descriptions (#13871) (`5df4cb3`) - fix: Refresh Factory workspaces from main (#13873) (`efa4ae0`) - perf: Speed up plain package listings (#13868) (`f68d214`) - fix: Detect package-level turbo.jsonc (#13875) (`e945403`) - fix: Format Slack pull request notifications (#13877) (`15c15f3`) - feat: Link workspaces to pull requests (#13878) (`ae25f97`) - chore: Upgrade repository to pnpm 12 (#13879) (`4e11bad`) - chore: Update with-vite-react example (#13880) (`54cf466`) - chore: Improve Factory chat streaming (#13881) (`eebf1b0`) - fix: Show workspace failure details (#13884) (`c7661b8`) - chore: Add selectable Factory coding harnesses (#13887) (`6b5ca16`) - fix: Gracefully handle native lockfile failures (#13886) (`463a292`) - chore: Update merged PR Slack messages (#13888) (`66ebcac`) - fix: Show native tasks in task listings (#13889) (`7e4d9d3`) - docs: Clarify turbo ls stability in Agent Skill (#13892) (`8ee758c`) - chore: Resume workspace chat on refocus (#13894) (`7026627`) - chore: Preserve HarnessAgent bridge assets (#13896) (`26f3160`) - fix: Deduplicate multi-language watch hashes (#13897) (`5ade89b`) - fix: Normalize multi-language cache environment inputs (#13898) (`cfca6f2`) - chore: Add workspace diff view (#13900) (`813d54a`) - refactor: Migrate CLI parser to usage-rs (#13890) (`712e0e4`) - fix: Update --filter help links to canonical docs URL (#13891) (`6cf116d`) - chore: Update with-vue-nuxt example (#13901) (`d10e3e0`) - docs: Document missing reference flags and fix query typo (#13902) (`1f2acc5`) - docs: Align CONTRIBUTING Node and pnpm versions (#13883) (`a3bda61`) - perf: Skip Cargo resolution for workspace-only queries (#13903) (`b928ecc`) - fix: Bound memory usage for streamed task output (#13908) (`7f85401`) - fix: Clarify generator action reporting (#13910) (`7fe0201`) - fix: Bump tar to 7.5.22 in @turbo/releaser (CVE-2026-73566) (#13911) (`aba64e9`) - fix: Correct devtools --port default in CLI help (#13905) (`11f7ee6`) - fix: Resolve lint errors (#13912) (`d06cf32`) - fix: Stabilize Rust builds on macOS (#13913) (`7d14df2`) - docs: Clean up Rust docs (#13914) (`3225cee`) - fix: Preserve task metadata for command overrides (#13915) (`54bbf8a`) - perf: Resolve Git SCM state in one process (#13882) (`7c993ef`) - fix: Only treat files literally named .gitignore as ignore files in the untracked walk (#13916) (`ac32a79`) - fix: Pass merge_base and allow_unknown_objects in signature order for task-level filter ranges (#13917) (`48f64e1`) - fix: Hash deferred inputs after dependencies (#13919) (`4cc80f7`) - test: Make Cargo hash fixtures hermetic (#13920) (`32ff2f3`) - fix: Allow release versioning with pending changes (#13921) (`5114901`) - chore: Update to Rust 1.98.0 (#13821) (`ebef942`) - chore: Upgrade Factory to Fable 5.1 (#13924) (`a57476f`) - docs: Align Python guide with Rust guide (#13927) (`f166370`) - fix: Give the package changes watcher the future flags so root inputs reach the task filter (#13926) (`e5ccf55`) - fix: Support pnpm 6 single-package lockfiles (#13928) (`a3d52a6`) - feat: Add skipPackageGraph option to @turbo/repository Workspace.find (#13929) (`99b5a00`) - fix: Use pnpm 12 in library release containers (#13930) (`0b6aaf9`) - fix: Upgrade repository N-API build tooling (#13931) (`dd6c1c6`) - fix: Use Node 20 for musl library builds (#13933) (`61b1635`) - fix: Bootstrap musl builds with sh (#13934) (`f9ac0c9`) - fix: Use static library bootstrap shell (#13935) (`f819dca`) - fix: Allow unclean tree in bump-version (#13936) (`28562e7`) - fix: Skip git commit in bump-version (#13937) (`ea4ba17`) - fix: Expect napi 3 artifact paths (#13938) (`6b5eea3`) - chore: Release Turbo repository packages 0.0.1-canary.27 (#13939) (`5c54e50`) - perf: Batch package detail queries (#13923) (`c8347b1`) - docs: Touch-ups for Python and Rust docs (#13940) (`f20b415`) - chore: Update basic example (#13922) (`b40dd07`) - fix: Avoid oxlint false positive on Workspace.find options (#13941) (`3125eb1`) - chore: Update kitchen-sink example (#13945) (`fd8dc59`) - fix: Hash native workspace task inputs (#13947) (`17d0940`) - chore: Update with-angular example (#13951) (`3770d4c`) - fix: Recompute eager hashing flag after prepending global inputs (#13949) (`9cea33b`) - refactor: Generalize daemon repository discovery (#13946) (`350bd41`) - feat: Make daemon discovery toolchain generic (#13952) (`d65c615`) - feat: Add experimental Go workspace graph foundation (#13953) (`6687fa8`) - feat: Add native Go task tables (#13954) (`dccc890`) - feat: Add Go task input and output contracts (#13955) (`8501ce7`) - test: Cover native Go execution end to end (#13956) (`48d4322`) - feat: Add Go external resolution fingerprints (#13957) (`f6d5f18`) - feat: Fingerprint Go toolchain environment (#13958) (`ec18363`) - feat: Add Go change observations (#13959) (`5256ecc`) - test: Cover Go query and summary surfaces (#13960) (`a24ea48`) - feat: Add Go-aware prune planning (#13961) (`026d7fb`) - test: Add end-to-end Go prune and cache coverage (#13962) (`a1cbfaa`) - docs: Document experimental native Go workspaces (#13963) (`7f36a42`) - fix: Harden Go workspace diagnostics (#13964) (`8a6b0f7`) - fix: Correct Go workspace resolution and task caching (#13966) (`8ef6aaf`) - chore: Add Factory thinking controls and message queue (#13967) (`77e975c`) - chore: Remove ARCHITECTURE.md docs (#13969) (`4f8e693`) - fix: Add slug to OPTIONS request (#13943) (`63797e4`) - perf: Parse CLI exits before starting worker pools (#13971) (`6888882`) - perf: Reuse compiled workspace matchers (#13970) (`7aaf7cb`) - chore: Revert bundled startup optimizations (#13982) (`c11e327`) - perf: Fast-path numeric package manager versions (#13973) (`3b28640`) - perf: Stop inference after selecting a workspace root (#13980) (`a3c5b9c`) - perf: Reuse compiled workspace matchers (#13984) (`eb80d6f`) - perf: Skip unnecessary glob normalization (#13979) (`1840e84`) - perf: Cache workspace configuration lookups (#13975) (`f59661a`) - perf: Avoid regexes for simple environment matching (#13974) (`7e5d793`) - perf: Reuse canonical paths during repository discovery (#13976) (`2ba3e6e`) - perf: Skip work for disabled telemetry (#13972) (`267ed73`) - perf: Defer filesystem checks for absent configs (#13977) (`f7db9df`) - perf: Compile workspace glob combinators directly (#13978) (`223fa9f`) - docs(skills): fix the dependencies-only filter caret position (#13965) (`73ac1fa`) - chore: Update with-biome example (#13983) (`bd331cf`) - fix: Normalize bare directory outputs when filtering dependencyOutputs hashes (#13932) (`555f2bc`) - chore: Scope panic lint allows in hash crates to tests and generated code (#13810) (`e901550`) - perf: Use binary search for resolution states (#13986) (`6a7f1c6`) - perf: Use dense membership for task graph pruning (#13987) (`4ffbaa9`) - perf: Throttle TUI tick cadence (#13988) (`b74be22`) - perf: Fast-path package-only filter selectors with --only (#13990) (`8059b91`) - perf: Avoid materializing npm's legacy dependency tree (#13993) (`acd35c8`) - perf: Stream custom repository downloads to disk (#13994) (`a6b88ee`) - perf: Copy shared file dependencies once per destination during prune (#13995) (`efbe558`) - perf: Index npm workspace links during subgraph extraction (#13997) (`996a0bb`) - perf: Cache Turbo config reads across ESLint rule invocations (#13998) (`c02f97e`) - perf: Scope untracked-file discovery to filtered runs (#13991) (`ead0e19`) - perf: Limit turbo.json preloading to required task scope (#13992) (`f788cd8`) - ci: Restrict release signing to main (#13985) (`2258541`) - perf: Coalesce glob-watcher invalidation snapshots (#13996) (`3fec2fe`) - perf: Reuse resolver inference during file tracing (#13999) (`596ea4c`) - chore: Release Turborepo 2.10.13-canary.2 (#14000) (`ebfc808`) - chore: Exclude turbo and @turbo/* from minimum release age (#14005) (`5897420`) - perf: Binary search boundaries comments before each import (#14003) (`6a95c81`) - perf: Fetch only npm dist tags in the binary version endpoint (#14007) (`4db9ca0`) - perf: Compute Devtools graph depths without rescanning every edge (#14001) (`ef46772`) - perf: Reuse package manager detection during create startup (#14006) (`06f5900`) - perf: Borrow workspace entries during dependency lookup (#14002) (`ba822e1`) - perf: Cache LSP task index and dedupe identities in constant time (#14008) (`77f6232`) - perf: Deduplicate recursive glob prefixes before walking (#14004) (`90fd2d6`) - perf: Batch boundaries progress updates (#14009) (`3862f3e`) - perf: Reuse archive anchor during cache creation (#14010) (`f01afa9`) - perf: Bound remote-cache transfer memory instead of retaining whole artifacts (#14011) (`470ffb9`) - perf: Build cache archives once for local and remote writes (#14012) (`f0a0d77`) - perf: Make VS Code binary discovery and LSP probes nonblocking (#14015) (`8cc0c42`) - perf: Avoid serializing ASTs for path-only file dependency queries (#14016) (`96712c9`) - perf: Share source text across boundaries diagnostics instead of copying it per error (#14014) (`5223bd6`) - perf: Parse the lockfile once when opening a native Workspace with its graph (#14017) (`23ee440`) - docs: Document a focused Cargo build for CLI-only development (#14018) (`d21277e`) - perf: Run the docs MDX generator once per quality task graph (#14019) (`f7b030b`) - perf: Keep package bundling inside a cacheable build boundary (#14020) (`f3ae66f`) - perf: Skip speculative content hashing for size-changed unnormalized tracked files (#14021) (`a8b4ac2`) - perf: Move synchronous cache archiving off Tokio runtime workers (#14013) (`22b6577`) - perf: Index package directories for changed-file ownership lookups (#14022) (`05f16eb`) - perf: Bound TUI raw-output retention independently of scrollback (#14024) (`0cb6d58`) - perf: Coalesce package rediscovery while a scan is already running (#14023) (`c6eac09`) - perf: Replace repeated deferred-hash scans with dependency-driven readiness (#14025) (`9a98e64`) - perf: Read child output in bounded chunks instead of per-line (#14026) (`f6214c8`) - perf: Share one repo index across same-package spec rehashes in watch mode (#14027) (`171a0a3`) - test: Deflake rediscovery_coalesces_during_in_flight_scan (#14029) (`d41369b`) - perf: Reuse the package graph across partial watch reruns (#14028) (`9a42fe3`) - test: Deflake Go watch CI (#14031) (`0af9b4f`) - chore: Release Turborepo 2.10.13-canary.3 (#14032) (`cfd82d2`) - docs: Group experimental languages and link the Go RFC (#14034) (`6e8f7fe`) - chore: Update with-docker example (#14035) (`25fe817`) - perf: Avoid cloning package list in turbo ls (#14036) (`f3b0fa9`) - fix: Restore leading flags for implicit run commands (#14037) (`3235ec5`) - fix: Include experimentalCI in task hashes (#14038) (`b8c7370`) - chore: Release Turborepo 2.10.13-canary.4 (#14039) (`17b358a`) - docs: Explain circular package dependencies in boundaries (#14042) (`895337e`) - docs: Document circular dependencies in boundaries (#14041) (`3821e42`) - fix: Preserve Yarn resolution declaration order (#14045) (`54b74a7`) - fix: Avoid duplicate interrupts during task shutdown (#14046) (`f9fa885`) - fix: Preserve nested Bun package keys (#14050) (`b647afc`) - chore: Update with-mcp-servers example (#14049) (`2c11c4f`) - chore: Release Turborepo 2.10.13-canary.5 (#14051) (`cc1caca`) - fix: Preserve native task dependencies across entrypoints (#14047) (`ce4f045`) - fix: Pass through LOCALAPPDATA by default (#14056) (`258a08d`) - chore: Release Turborepo 2.10.13-canary.6 (#14057) (`bea9b6c`) - docs: Fix some typos (#14058) (`6a3b4c1`) - feat: Add Remix inference and `NUXT_ENV_*` wildcard (#14060) (`2167e74`) - fix: Return resolved package names for aliases (#14061) (`83f3ee1`) - chore: Release Turbo repository packages 0.0.1-canary.28 (#14062) (`e5094c7`) - feat: Add NEXT_DEPLOYMENT_ID to Next.js framework inference (#14065) (`36383cb`) - chore: Release Turborepo 2.10.13 (#14066) (`05a1d4b`) - fix: Fall back to the published tag when the dogfooded turbo version is unpublished (#14067) (`4d13cb5`) - feat: Expose compact task planning through turbo query (#14068) (`1e9336e`) - chore: Release Turborepo 2.10.14-canary.1 (#14069) (`9077648`) - fix: Discover native workspace metadata lazily (#14053) (`30e0250`) - fix: Run Go workspace formatting per module (#14070) (`81d62ff`) - fix: Run Go verification per module (#14072) (`e1eb388`) - fix: Use pnpm for non-monorepo deployments (#14071) (`53dadbe`) - chore: Update with-nestjs example (#14064) (`a534c7e`) - docs: Document per-toolchain task commands (#14073) (`bc9017a`) - fix: Infer all co-located packages from cwd (#14074) (`46d0fd6`) - fix: Preserve all co-located package change owners (#14075) (`064d74a`) - fix: Isolate task logs for co-located packages (#14077) (`16b312f`) - chore: Make filesystem watch cleanup idempotent (#14078) (`441f205`) - fix: Follow Go default binary output paths (#14079) (`a1ab64a`) - chore: Release Turborepo 2.10.14-canary.2 (#14080) (`c2ddd13`) - feat: Make package queries and pruning task-aware (#14081) (`012b091`) - chore: Release Turborepo 2.10.14-canary.3 (#14082) (`7b2abd7`) - chore: Update with-npm example (#14083) (`d7f89ba`) - feat: Expose static repository discovery and safe affected candidates (#14085) (`22a201d`) - fix: Preserve task env mode when pruning (#14087) (`7eda8d5`) - feat: Derive short names for Go workspace packages (#14086) (`b3cfa86`) - chore: Release Turbo repository packages 0.0.1-canary.29 (#14088) (`15830dc`) - chore: Release Turborepo 2.10.14-canary.4 (#14089) (`b66c30d`) - feat: Infer Cargo and uv affectedness from static package inputs (#14090) (`728a79d`) - chore: Release Turbo repository packages 0.0.1-canary.30 (#14091) (`438c45b`) - perf: Cache Windows archive anchor (#14093) (`0c3bc42`) - fix: Harden cache archive restoration (#14095) (`7e7ab44`) - fix: Defer dependency output validation for lazy scopes (#14097) (`d804909`) - chore: Release Turborepo 2.10.14-canary.5 (#14098) (`613b3ba`) - test: Fix Windows cache restore test (#14099) (`2c49c23`) - chore: Update with-prisma example (#14101) (`f21d73f`) --------- Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>

vercel Bot deployed to Preview – turborepo-factory September 7, 2026 23:36 View deployment

vercel Bot deployed to Preview – turborepo-factory September 8, 2026 00:32 View deployment

vercel Bot deployed to Preview – turborepo-factory September 8, 2026 03:17 View deployment

anthonyshew added 11 commits September 8, 2026 05:50

perf: Parse CLI exits before starting worker pools

4e99d08

perf: Skip work for disabled telemetry

2e49ecc

perf: Fast-path numeric package manager versions

a4ba878

perf: Avoid regexes for simple environment matching

6998657

perf: Cache workspace configuration lookups

30254c8

perf: Reuse canonical paths during repository discovery

6422b7a

perf: Defer filesystem checks for absent configs

6d00516

perf: Compile workspace glob combinators directly

1960dfa

perf: Skip unnecessary glob normalization

c68348f

perf: Stop inference after selecting a workspace root

bc3e08c

perf: Reuse compiled workspace matchers

d2fa588

anthonyshew force-pushed the shew/reduce-cli-startup-overhead branch from 6fc528b to d2fa588 Compare September 8, 2026 11:54

vercel Bot temporarily deployed to Preview – turborepo-factory September 8, 2026 11:55 Inactive

anthonyshew changed the title ~~perf: Reduce CLI startup overhead~~ perf: Reuse compiled workspace matchers Sep 8, 2026

anthonyshew mentioned this pull request Sep 8, 2026

perf: Stop inference after selecting a workspace root #13980

Merged

anthonyshew changed the base branch from main to shew/startup-10-root-inference September 8, 2026 11:56

anthonyshew requested a review from a team as a code owner September 8, 2026 12:16

anthonyshew requested review from tknickman and removed request for a team September 8, 2026 12:16

anthonyshew changed the base branch from shew/startup-10-root-inference to main September 8, 2026 12:25

anthonyshew requested a review from a team September 8, 2026 12:25

Merge branch 'main' into shew/reduce-cli-startup-overhead

2405ca4

vercel Bot deployed to Preview – turborepo-factory September 8, 2026 12:28 View deployment

test: Stabilize Rust CI checks

9c152aa

vercel Bot deployed to Preview – turborepo-factory September 8, 2026 12:45 View deployment

anthonyshew merged commit 7aaf7cb into main Sep 8, 2026
49 of 50 checks passed

anthonyshew deleted the shew/reduce-cli-startup-overhead branch September 8, 2026 12:59

anthonyshew mentioned this pull request Sep 8, 2026

chore: Revert bundled startup optimizations #13982

Merged

anthonyshew mentioned this pull request Sep 8, 2026

perf: Reuse compiled workspace matchers #13984

Merged

github-actions Bot mentioned this pull request Sep 9, 2026

chore: Release Turborepo 2.10.13-canary.2 #14000

Merged

github-actions Bot mentioned this pull request Sep 14, 2026

chore: Release Turborepo 2.10.13 #14066

Merged

github-actions Bot mentioned this pull request Sep 18, 2026

chore: Release Turborepo 2.11.0 #14102

Merged

Navigation Menu

perf: Reuse compiled workspace matchers - #13970

perf: Reuse compiled workspace matchers#13970
anthonyshew merged 13 commits into
mainfrom
shew/reduce-cli-startup-overhead

anthonyshew commented Sep 7, 2026 •

edited

Loading

Uh oh!

vercel Bot commented Sep 7, 2026 •

edited

Loading

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Conversation

anthonyshew commented Sep 7, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Description

Testing Instructions

Measurement limits

Uh oh!

vercel Bot commented Sep 7, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

anthonyshew commented Sep 7, 2026 •

edited

Loading

vercel Bot commented Sep 7, 2026 •

edited

Loading