1. 25 Sep, 2026 38 commits
    • next-js-bot[bot]'s avatar
      v16.4.0-canary.48 · ff6ac752
      next-js-bot[bot] authored
      ff6ac752
    • Josh Story's avatar
      Hoist resolved pathname into app page route match (#99259) · 2ebafcbc
      Josh Story authored
      This PR was authored by codex but I'll share the motivation here in my
      own words. In an effort to improve the architectural underpinnings of
      Next.js we need to move away from a request/response dominated interface
      for rendering in Next.js. The legacy of this style is code that in some
      cases targets a fake Request or Response object even in cases where we
      are prerendering and no such objects actually exist. This PR adds a new
      concept of RouteMatch which seems overkill for only containing the
      resolvedPathname however we will be migrating more properties out of
      catchall context such as renderOpts and other request local context into
      more semantically meaningful packaging.
      
      ## Summary
      
      Introduce a required `RouteMatch` containing the route-prepared
      `resolvedPathname` and pass it explicitly through the Node, Edge, direct
      server, and static export app-page render entrypoints.
      
      The type is owned by the app-page route-module handler boundary rather
      than app-render. `resolvedPathname` is no longer written to or read from
      request metadata, preventing downstream code with request access from
      depending on that side channel while preserving the existing
      delimiter-safe pathname representation.
      
      ## Verification
      
      - `pnpm prettier --with-node-modules --ignore-path .prettierignore
      --write <changed files>`
      - `npx eslint --config eslint.config.mjs --fix <changed files>`
      - `git diff --cached --check`
      - Source grep confirming there are no `resolvedPathname`
      request-metadata reads, writes, or fields
      - Could not complete: `pnpm --filter=next types` (this fresh worktree
      lacks per-package dependencies and compiled declarations; the check
      failed broadly on unrelated missing modules)
      
      <!-- NEXT_JS_LLM -->
      2ebafcbc
    • Joseph's avatar
      docs: papercuts in image API reference (#99256) · 779e73eb
      Joseph authored
      - The `sizes` section didn't explain what happens to `fill` images
      without `sizes`. Only images with a `width` get the limited `1x`/`2x`
      `srcset`
         - `fill` images get one built from `deviceSizes`.
      - Updated `deviceSizes` to match: it's also used for `fill` images
      without `sizes`.
      
      - "Forever" isn't an accurate description of static image import
      caching. The built-in image optimizer serves these with `public,
      max-age=315360000, immutable`, which is 10 years
      ([source](https://github.com/vercel/next.js/blob/aea417ab0c490ec3e07112ed48ba20236fdfad18/packages/next/src/server/image-optimizer.ts#L305-L309),
      [header](https://github.com/vercel/next.js/blob/aea417ab0c490ec3e07112ed48ba20236fdfad18/packages/next/src/server/image-optimizer.ts#L831)).
      - Changed to "long-lived `immutable`" in the image and `legacy` image
      references.
      779e73eb
    • Jamiboy Mohammad's avatar
      test: run the TypeScript version warning suites in deploy (#99195) · 2f3842dd
      Jamiboy Mohammad authored
      ## Summary
      
      Both suites assert on build output — one that `next build` warns about
      TypeScript 4.0.6, the other that it stays quiet on a current version.
      Deploy captures build output, so there is nothing here a deployment
      cannot observe.
      
      They were excluded by `skipDeployment` plus an `if (isNextDeploy ||
      isNextDev)` return that reads as caution rather than a finding: no
      incompatibility is recorded anywhere, and neither suite touches the
      filesystem or the local CLI.
      
      ```diff
      -  const { next, isNextDeploy, isNextDev, skipped } = nextTestSetup({
      +  const { next, isNextDev } = nextTestSetup({
           files: __dirname,
           skipStart: true,
      -    skipDeployment: true,
      -  if (skipped) { return }
      -  if (isNextDeploy || isNextDev) {
      +  if (isNextDev) {
           it('should skip', () => {})
           return
         }
      ```
      
      The `isNextDev` half of the guard is the part that holds — the warning
      is emitted during a build, and there is no build in dev.
      
      ## The 240s timeout is not incidental
      
      Each suite calls `next.start()` **inside the test body**, which
      `e2e-utils` caps at 60 seconds when no explicit timeout is given.
      Deployments now take about two minutes, so without this both would time
      out on every run — the same failure #99162 had to fix in
      `deprecation-warnings`.
      
      ## Verification
      
      Run against real deployments in deploy mode with **webpack**, which is
      the bundler PR CI does not cover:
      
      ```
      PASS webpack  typescript-version-warning      57.7s  ✓
      PASS webpack  typescript-version-no-warning   81.6s  ✓
      ```
      
      Also green in start mode with Turbopack. ESLint and Prettier clean.
      
      ## Two suites I deliberately left out
      
      These came from the same review of 19 suites that reported green in
      ~2.4s without running anything:
      
      - **`url-imports`** is correctly excluded. It starts a local static
      server on port 12345 and imports from it, so no deployment can reach its
      source regardless of bundler. Its `skipDeployment` is load-bearing.
      - **`webpack-loader-parse-error`** is already being enabled in #98850.
      
      🤖 Generated with [Claude Code](https://claude.com/claude-code)
      
      ---------
      Co-authored-by: 's avatarClaude Opus 5 <noreply@anthropic.com>
      2f3842dd
    • Jamiboy Mohammad's avatar
      test: drop skipDeployment where another gate already excludes deploy (#99194) · 4562ce8c
      Jamiboy Mohammad authored
      ## Summary
      
      Eleven suites carry `skipDeployment` on top of a condition that already
      keeps them out of deploy mode, so the option is dead weight.
      
      Each was run in deploy on the #99184–#99187 batches with the exclusion
      removed, and each reported green in roughly **2.4s** — the signature of
      a suite that registered no tests, not one that deployed and passed. So
      removing the option is behaviour-neutral.
      
      It also removes the `it.only` that `skipDeployment` registers, which
      focuses the whole *file* and can silently skip unrelated suites beside
      it.
      
      ## Why each genuinely cannot run against a deployment
      
      | suite | blocker |
      |---|---|
      | `app-action-export`, `invalid-custom-routes`,
      `invalid-middleware-matchers` | mutate fixture files after setup
      (`patchFile` / `deleteFile`), then rebuild |
      | `instant-validation-build`, `externals-pages-bundle` | call
      `next.build()` and read the local build output |
      | `telemetry/config`, `telemetry/page-features`, `telemetry/telemetry` |
      telemetry events are emitted by the local CLI process |
      | `cpu-profiling-build` | reads `.next-profiles` off disk, and already
      returns early on `isNextDeploy` |
      | `back-button-download-bug`, `next-test` | the whole suite is
      `describe.skip`, for reasons unrelated to deployment |
      
      ## Not in this PR
      
      From the same set of 19, four look like the *gating* is wrong rather
      than the exclusion being correct, and they need their own change plus a
      deploy run:
      
      - `typescript-version-warning` and `typescript-version-no-warning` use
      only `next.start` and `next.cliOutput`, and assert on build output that
      deploy does provide. They are excluded by an explicit `if (isNextDeploy
      || isNextDev)` that reads as caution rather than a finding.
      - `url-imports` uses only `render` and `browser`. Its `describe.skip` is
      Turbopack-specific and legitimate, so it needs the webpack deploy job
      rather than a code change.
      - `webpack-loader-parse-error` is already being enabled in #98850.
      
      And `invalid-module-document`, `undefined-webpack-config`,
      `typegen-error-diagnostic`, `turbopack-emit-collect` need the
      deferred-deployment-lifecycle conversion or a closer read first.
      
      ## Verification
      
      Deploy-mode runs on #99184–#99187 (Turbopack). Confirmed per file that
      canary carries one `skipDeployment` and this branch carries none, with
      no other change.
      
      🤖 Generated with [Claude Code](https://claude.com/claude-code)
      
      ---------
      Co-authored-by: 's avatarClaude Opus 5 <noreply@anthropic.com>
      4562ce8c
    • Will Binns-Smith's avatar
      bundle-analyzer: suspend while loading analyzer data (#98535) · 2757d564
      Will Binns-Smith authored
      ## Summary
      
      Use SWR Suspense for required analyzer datasets and move loading
      behavior into explicit Suspense and error boundaries.
      
      Split selected-route and valid-comparison subtrees so resolved route and
      module data has non-null contracts. Keep optional history, metadata, and
      one-sided comparison data nullable where absence is part of the UI
      state.
      
      
      ## Verification
      
      - `pnpm --filter=@next/bundle-analyzer-ui exec tsc --noEmit`
      - focused ESLint for changed files
      - `pnpm --filter=@next/bundle-analyzer-ui build`
      
      <!-- NEXT_JS_LLM -->
      
      ---------
      Co-authored-by: 's avatarvercel-fleet-prod[bot] <318278635+vercel-fleet-prod[bot]@users.noreply.github.com>
      Co-authored-by: 's avatarWill Binns-Smith <755844+wbinnssmith@users.noreply.github.com>
      2757d564
    • Hendrik Liebau's avatar
      Fix Server Actions hanging after navigation to PPR pages (#99252) · 6137b367
      Hendrik Liebau authored
      Adopts #99229. Closes #99229.
      
      A Server Action can hang on Vercel when the client navigates from `/a`
      to a PPR `/b` page and then calls an action owned by `/a`. The `/b` POST
      can carry postponed state. `app-page-runtime.ts` consumes the request
      stream to separate that state and saves the remaining action body as
      `actionBody`. The forwarding path reads `req.stream()` again, so its
      fetch waits on an exhausted stream until the function times out.
      
      The forwarding path now sends `actionBody` when it is present. It
      removes the source route's resume headers because the forwarded body no
      longer contains that state; the platform can attach the target route's
      state on the new request. Requests without a saved action body still use
      the request stream.
      
      The E2E test uses `connection()` to make `/b` PPR, warms its shell,
      navigates from `/a` to `/b`, and invokes an action imported only by
      `/a`. On a fresh Vercel deployment, it stays pending without this change
      and returns `pong` with it. The same test passes under `next start`,
      which does not receive the platform's prepended state.
      
      Fixes #99153
      
      ---------
      Co-authored-by: 's avatarDmitrii Troitskii <jsleitor@gmail.com>
      6137b367
    • Will Binns-Smith's avatar
      Fix blurry bundle analyzer treemap on low-DPI displays (#99197) · 36e51187
      Will Binns-Smith authored
      ## Summary
      
      Size the treemap canvas from its content box so the backing store is not
      scaled to fit inside the bordered container. Recreate the backing store
      when the device pixel ratio changes.
      
      ## Verification
      
      - Manually verified treemap rendering and backing dimensions on
      standard- and high-DPI displays
      
      <!-- NEXT_JS_LLM -->
      36e51187
    • Will Binns-Smith's avatar
      Rename bundle analyzer snapshot naming option (#99251) · a0ed5fa7
      Will Binns-Smith authored
      ### What?
      Rename the Turbopack bundle analyzer option to `--snapshot-name` and use
      `snapshotName` in saved snapshot metadata and the comparison UI. Add an
      E2E regression test for live and historical metadata.
      
      ### Why?
      The custom label names a snapshot, not just a comparison baseline; the
      old term was misleading in the CLI and saved data.
      
      ### How?
      Carry the new option through the analyze pipeline into the snapshot
      metadata used by both live and historical views. The UI uses that
      metadata for display labels and picker search. The old CLI option and
      metadata field are not retained; existing snapshots containing only
      `baselineName` no longer show their custom label.
      
      ### Verification
      - Static wiring checks and `git diff --check` passed.
      - E2E test could not run locally: dependencies are not installed (`jest:
      not found`).
      
      <!-- NEXT_JS_LLM -->
      
      <!-- fleet ceb93e13-30fe-4be3-a35e-c7282c730c3d -->
      
      ---------
      Co-authored-by: 's avatarvercel-fleet-prod[bot] <318278635+vercel-fleet-prod[bot]@users.noreply.github.com>
      Co-authored-by: 's avatarWill Binns-Smith <755844+wbinnssmith@users.noreply.github.com>
      a0ed5fa7
    • Tobias Koppers's avatar
      Reduce stacked PR gate polling and use ts-jest (#99215) · 4c055105
      Tobias Koppers authored
      ## Summary
      
      Follow-up to merged #99086, addressing [bgw’s Jest transformer
      review](https://github.com/vercel/next.js/pull/99086#discussion_r4100921785)
      and the remaining GitHub API rate-limit concern. The gate still waits
      for **any of the nearest three PRs** to pass its required check, failing
      closed if all three fail. Fork, leaf and first-three PR behavior is
      unchanged.
      
      - Refresh the current PR / branch-link topology every **20 minutes**
      while waiting; on each intervening **five-minute** poll, fetch only the
      up-to-three predecessor check statuses. Revalidate live PR revisions and
      checks before *any* success/failure decision. A changed decisive
      snapshot triggers an immediate topology refresh; repeated invalidation
      backs off.
      - Replace the bespoke TypeScript Jest transformer with `ts-jest`, using
      the action’s local pinned Jest 29 runner and peers. Keep independent
      `tsc --noEmit` typechecking and TypeScript 6.0.2. Refresh the compiled
      action and fake-timer tests.
      
      ### REST request rate per steadily waiting middle PR
      
      | Poll | Previous | New |
      | --- | ---: | ---: |
      | First poll and every 20 minutes | 8 | 8 (topology + up to 3 checks) |
      | 5/10/15 minutes between topology refreshes | 8 each | 3 each (checks
      only) |
      | **20-minute total** | **32** | **17** |
      | **Steady-state hourly rate** | **96/hour** | **51/hour (~47% fewer)**
      |
      
      This assumes one page per PR list, three predecessors and an unresolved
      wait. It is **not a hard per-PR or repo-wide ceiling**: a cold poll uses
      eight reads, decision-time live revalidation adds up to **3** (success)
      or **7** (all-failed) reads, and pagination, transient retries or an
      early topology refresh after changed data add calls. Changes that do not
      coincide with a decisive check can take up to 20 minutes to be detected;
      no cached state alone may release or fail the required gate.
      
      **Dependency caveat:** `ts-jest@29.4.5` declares a TypeScript `<6` peer
      range, while this repository pins TS 6.0.2. pnpm warns when adding the
      dependency; the separate action workspace pins Jest and its `ts-jest`
      peers to 29.x, and both the actual 27-case Jest suite and independent
      TypeScript check pass without suppressing errors.
      
      ## Verification
      
      - `pnpm --dir .github/actions/pr-stack-ci-gate install --frozen-lockfile
      --ignore-scripts`
      - `pnpm --dir .github/actions/pr-stack-ci-gate types`
      - `pnpm --dir .github/actions/pr-stack-ci-gate test` — **27/27** focused
      tests, including exact t=0/5/10/15/20 REST-call timings, stale-branch
      revalidation, a same-SHA rerun and a transient status-poll outage.
      - `pnpm --dir .github/actions/pr-stack-ci-gate build` — committed ncc
      bundle is reproducible; workflow YAML parses and the six-hour gate
      timeout remains.
      - `pnpm lint-no-typescript`, Prettier and `git diff --check`.
      - Live PR CI on signed head `1c0feed844b598b84736a849f72d4af7f2ff9831`:
      [compiled
      gate](https://github.com/vercel/next.js/actions/runs/36105343891/job/107976526834)
      **passed**, and
      [lint](https://github.com/vercel/next.js/actions/runs/36105343891/job/107976745858)
      **passed**, including frozen install, typecheck, bundle build and
      **27/27 Jest tests**. The [full build-and-test
      workflow](https://github.com/vercel/next.js/actions/runs/36105343891)
      completed **successfully on attempt 1**, including the required [`thank
      you,
      next`](https://github.com/vercel/next.js/actions/runs/36105343891/job/107984156677)
      check. A normal top/leaf PR runs immediately; its CI cannot exercise a
      20-minute middle-PR wait. No six-PR draft-stack rebase is included in
      this follow-up.
      - README-wording-only follow-up head
      `f7d58fa85338f38b1a9876e308edb613104c5e8e`: [new build-and-test
      run](https://github.com/vercel/next.js/actions/runs/36184841154)
      **queued** at the time of this update. The previous full-CI pass belongs
      to `1c0feed8`, not this new SHA; no executable files changed.
      
      <!-- NEXT_JS_LLM -->
      
      
      <!-- fleet d2ae64f1-7106-4ba4-b88a-7785c19904c6 -->
      
      ---------
      Co-authored-by: 's avatarvercel-fleet-prod[bot] <318278635+vercel-fleet-prod[bot]@users.noreply.github.com>
      Co-authored-by: 's avatarTobias Koppers <1365881+sokra@users.noreply.github.com>
      4c055105
    • Luke Sandberg's avatar
      Name the internal trace server process (#99257) · 46272dcf
      Luke Sandberg authored
      ## Summary
      
      Give the long-running `next internal trace` server a recognizable
      process title with the Next.js version, matching the version format used
      by the build and server commands. Set the title when the trace command
      starts so both the WebSocket and MCP services appear under the same
      name, without changing their behavior.
      
      This way that `node` process consuming 40g of ram is a little easier to
      understand
      <!-- NEXT_JS_LLM -->
      
      <!-- fleet 1cc505a4-180a-44ae-8200-3af268332f61 -->
      Co-authored-by: 's avatarvercel-fleet-prod[bot] <318278635+vercel-fleet-prod[bot]@users.noreply.github.com>
      Co-authored-by: 's avatarLuke Sandberg <210140+lukesandberg@users.noreply.github.com>
      46272dcf
    • Jamiboy Mohammad's avatar
      test: enable deploy coverage for asset and compiler errors (#98850) · d3499931
      Jamiboy Mohammad authored
      ## Summary
      Enable deployment coverage for 15 asset/compiler exclusion locations
      using the deferred deployment lifecycle from #98776. Replace local-only
      build calls with explicit start/failure assertions and retain the
      expected image, Sass, loader, foreign-package, and missing-dependency
      diagnostics. Preserve the existing development-only and bundler-specific
      exclusions.
      
      ## Verification
      - All applicable focused Turbopack production/development and Webpack
      production checks passed.
      - Local checks used the published native binding. Exact native and
      deploy validation are pending repository CI.
      
      <!-- NEXT_JS_LLM -->
      
      ---------
      Co-authored-by: 's avatarClaude Opus 5 <noreply@anthropic.com>
      d3499931
    • Jamiboy Mohammad's avatar
      test: consolidate duplicate mode branches in build error tests (#99135) · ef0d79f5
      Jamiboy Mohammad authored
      ## Summary
      
      Follow-up to #98847 and #98848, addressing [this review
      comment](https://github.com/vercel/next.js/pull/98847#discussion_r_placeholder)
      about the duplicate `isNextDev` check in
      `cache-components-edge-deduplication`.
      
      Several suites that gained deploy coverage branch on the test mode twice
      in a row: once to decide whether `next.start()` should resolve or
      reject, and again to pick the assertions. Each pair is folded into a
      single branch so the start call sits with the assertions that depend on
      it. Two suites that wrote `if (isNextDev) { ... }` followed by `if
      (!isNextDev) { ... }` now use a plain `else`, and
      `interception-routes-output-export` drops an `else if (isNextDev)` that
      could only ever be an `else`.
      
      No assertion changes.
      
      - `cache-components-edge-deduplication`
      - `cache-components-route-handler-errors`
      - `app-edge-invalid-reexport` — the start branch also folded the
      `isTurbopack` distinction in, so the "Turbopack fails, webpack warns"
      split is stated once
      - `generate-static-params-error` — the dev-only `beforeAll` moved into
      the dev branch
      - `interception-routes-output-export`
      - `catches-missing-getStaticProps`
      - `conflicting-public-file-page`
      
      ## Verification
      
      - `pnpm test-start-turbo` — 7/7 suites pass
      - `pnpm test-dev-turbo` — 7/7 suites pass
      - `pnpm test-start-webpack` — 6 pass, 1 skipped
      (`cache-components-edge-deduplication` is `@force-gate turbopack`)
      - `pnpm test-dev-webpack` — 6/6 applicable suites pass
      
      Local checks used the published native binding. Deploy validation is
      pending repository CI.
      
      🤖 Generated with [Claude Code](https://claude.com/claude-code)
      Co-authored-by: 's avatarClaude Opus 5 <noreply@anthropic.com>
      ef0d79f5
    • Jamiboy Mohammad's avatar
      test: snapshot the proxy-runtime build error output (#99136) · 4b9d9de2
      Jamiboy Mohammad authored
      ## Summary
      
      Follow-up to #98848, addressing [this review
      comment](https://github.com/vercel/next.js/pull/98848#discussion_r4074331118)
      asking for a snapshot test.
      
      The Turbopack production assertion in `proxy-runtime` compared a
      hand-maintained `expected` literal against the CLI output with
      `toContain`, with the deploy differences patched up by a local
      `normalize` helper. It now extracts the build error section and asserts
      it with an inline snapshot, so the whole diagnostic is pinned —
      including the error count line and the trailing frame, which the
      substring match never covered — and can be regenerated instead of edited
      by hand.
      
      The normalization deploy needed (timestamp prefixes, code-frame
      indentation, dropped blank lines) runs in every mode now, so a single
      snapshot covers local and deploy builds rather than branching on
      `isNextDeploy`.
      
      ### Tests that were checked and left alone
      
      `use-cache-close-over-function` and `use-cache-hanging` have the same
      `expected` + `normalize` + `toContain` shape from #98847, but their
      build errors happen during page generation, so the surrounding output
      interleaves `Generating static pages using N workers (n/m)` lines — `N`
      follows the host's CPU count and the progress lines vary run to run. In
      `use-cache-close-over-function` the failing route also varies, which is
      why the existing assertion matches `"/(client|server)"`. A snapshot over
      that region would be flaky, so both keep their targeted substring
      assertions. `proxy-runtime` is snapshot-safe because its error is a
      compile error raised before page generation starts.
      
      ## Verification
      
      - `pnpm test-start-turbo` — passes, and a second run passes against the
      committed snapshot
      - `pnpm test-start-webpack`, `pnpm test-dev-turbo`, `pnpm
      test-dev-webpack` — pass (these take the unchanged non-Turbopack/dev
      branch)
      
      Local checks used the published native binding. Deploy validation is
      pending repository CI.
      
      🤖 Generated with [Claude Code](https://claude.com/claude-code)
      Co-authored-by: 's avatarClaude Opus 5 <noreply@anthropic.com>
      4b9d9de2
    • Will Binns-Smith's avatar
      Avoid obsolete pnpm overrides in pack-next (#99199) · aea417ab
      Will Binns-Smith authored
      ### What?
      
      Make `pack-next` choose pnpm override locations from the destination
      workspace root's pinned `packageManager` version.
      
      ### Why?
      
      pnpm 11+ warns about `pnpm.overrides` in `package.json` and ignores that
      setting. Conversely, rewriting `pnpm-workspace.yaml` for older pnpm
      versions creates unnecessary churn.
      
      ### How?
      
      Use the repository's `semver` package to read the major from a pinned
      pnpm version. For a pnpm 11+ destination, merge overrides into the
      existing/detected pnpm workspace file without adding `pnpm.overrides` to
      `package.json`. For pnpm 10 and older, update only the package.json
      field. When the pin is missing or unusable, retain the previous
      dual-write behavior. Existing user overrides and npm/Yarn override
      behavior are preserved.
      
      ### Verification
      
      - Passed eleven temporary destination scenarios covering pinned
      older/newer pnpm, existing overrides and workspace files, nested
      workspace roots, and unknown versions.
      - Prettier check, Node syntax check, and `git diff --check` passed.
      - Repo-wide lint/typecheck not run (root dependencies are not installed
      in this checkout).
      
      <!-- NEXT_JS_LLM -->
      
      
      <!-- fleet e86eb7e1-c05e-4e30-9285-7c5595ccb36f -->
      
      ---------
      Co-authored-by: 's avatarvercel-fleet-prod[bot] <318278635+vercel-fleet-prod[bot]@users.noreply.github.com>
      Co-authored-by: 's avatarWill Binns-Smith <755844+wbinnssmith@users.noreply.github.com>
      aea417ab
    • Will Binns-Smith's avatar
      bundle-analyzer: move comparison state into URL (#98530) · bd6bdd1a
      Will Binns-Smith authored
      ## Summary
      
      Move bundle comparison into a dedicated `/compare` route and persist the
      selected baseline, comparison snapshot, route, and view in URL search
      parameters. This makes comparison views linkable and separates the App
      Router route entries from the shared client implementation.
      
      Use root-relative data URLs and proxy historical snapshots during
      development so analyzer data continues to load from nested routes.
      
      Follow-up to review feedback on #93521.
      
      ## Verification
      
      - `pnpm --filter=@next/bundle-analyzer-ui build`
      
      <!-- NEXT_JS_LLM -->
      bd6bdd1a
    • Luke Sandberg's avatar
      simplify `Transition` by removing dead members (#94467) · 91d6d671
      Luke Sandberg authored
      Remove a number of functions from `Transition` that are never overridden
      and had trivial implementations.
      
      This reveals that the default `process_context` function is making a
      copy of the module context simply without the `transition` (makes sense)
      but also with the `replace_externals` property reset to `true`. It isn't
      clear if that is intentional?
      
      This is just a code health thing no real perf win expected (save a tiny
      bit of code size)
      91d6d671
    • Jamiboy Mohammad's avatar
      fix: respect metadata streaming policy in Edge SSR (#99128) · b67a1182
      Jamiboy Mohammad authored
      ## Summary
      
      Edge App Router SSR always enabled streaming metadata, so requests
      matching the default HTML-limited bot list could receive an initial
      `<head>` without their delayed metadata. Use the same user-agent-based
      metadata policy as Node SSR so matching requests receive their metadata
      before `</head>`.
      
      Add regression coverage asserting the default streaming and blocking
      behavior in both Edge and Node runtimes. Twitterbot and TelegramBot
      receive their metadata in the initial head; ChatGPT-User, GPTBot,
      ClaudeBot, and browser requests continue to stream metadata. Also verify
      that delayed metadata eventually arrives in both runtimes. The default
      bot list is unchanged, and the tests do not override `htmlLimitedBots`.
      
      Related to #98748; this addresses the Edge runtime bug without making
      assumptions about how these bots consume streamed responses.
      
      ## Verification
      
      - Production regression suite: 14 tests passed with Webpack and 14 with
      Turbopack.
      - Full native bootstrap failed at `cargo metadata`; JavaScript build
      succeeded. Tests used the installed SWC 16.4.0-canary.38 binary with
      Next.js 16.4.0-canary.40.
      
      <!-- NEXT_JS_LLM -->
      b67a1182
    • next-js-bot[bot]'s avatar
      v16.4.0-canary.47 · 2b776990
      next-js-bot[bot] authored
      2b776990
    • Hendrik Liebau's avatar
      Prepare `'use cache'` for static root-param tracking (#99239) · fae4e1dd
      Hendrik Liebau authored
      Static root-param tracking for `'use cache'` needs to distinguish which
      getters a cached module imports. The existing generated module contains
      every getter, so module-level dependency collection includes unrelated
      root params.
      
      `next/root-params` now re-exports each getter from a separate virtual
      module. The generated `next/root-params` module declares that it has no
      side effects, so Turbopack resolves named imports without including
      unrelated getter modules. Runtime root-param tracking remains unchanged.
      
      The split also keeps code hashes stable when unrelated root params are
      added. The regression test checks this dependency isolation through the
      existing code-hash manifest.
      fae4e1dd
    • Sebastian "Sebbie" Silbermann's avatar
      [test] Adjust the prerender-encoding open fallback gate for non-adapter deploys (#99240) · 79647071
      Sebastian "Sebbie" Silbermann authored
      Test passes without adapter:
      https://github.com/vercel/next.js/actions/runs/36130462440/attempts/2
      
      Forked assertions with a clear statement what the intended behavior is
      would be more helpful.
      Co-authored-by: 's avatarClaude Code (kimi-k3[1m]) <noreply@anthropic.com>
      79647071
    • Tobias Koppers's avatar
      Turbopack: avoid project-wide traces from package directories (#99238) · 08aab14b
      Tobias Koppers authored
      ### What?
      
      Limit Turbopack's package-origin project-root tracing to statically
      named **files**, while resolving the complete original request relative
      to the package as before. This applies to filesystem, directory,
      child-process and Node worker references.
      
      ### Why?
      
      Static does not necessarily mean narrow: `node-gyp-build`'s
      `path.resolve(dir || '.')` contains a static `'.'` branch. Tracing that
      from the application root as a directory recursively includes unrelated
      application assets. We still need packages to find explicitly named
      application config files.
      
      ### How?
      
      Analysis records two candidate contexts without inspecting the
      filesystem: an optional app-root context for static file paths, and a
      package-local fallback for the original request. At reference
      resolution, the app-side lookup keeps only fully static file matches;
      directories and symlinked directories cannot recursively expand there.
      The package-side lookup independently traces the full pattern, including
      dynamic alternatives and directory contents. Both results are combined
      so even a mixed file/directory expression can retain its app config
      **and** package-local assets; the combination preserves distinct files
      sharing a request key and deduplicates identical modules. Protobuf
      `includeDirs` uses the same reference path.
      
      Regression fixtures cover a `node-gyp-build`-style `'.'` fallback, mixed
      file/directory alternatives with same-named files in both contexts,
      symlinked directories and files, a missing optional config, and Node
      worker references. The webpack/@vercel/nft path is unchanged.
      
      **Verification:** 93 tracing tests and 65 core resolve tests pass;
      focused Clippy with warnings denied, Rust formatting and diff checks
      pass. Production next-site size impact has not been independently
      measured in this fixture.
      
      Related: #98628, #98712, #99080.
      
      <!-- NEXT_JS_LLM -->
      
      
      <!-- fleet fe000a60-fbb8-4835-b6ef-38d5684822a9 -->
      
      ---------
      Co-authored-by: 's avatarvercel-fleet-prod[bot] <318278635+vercel-fleet-prod[bot]@users.noreply.github.com>
      Co-authored-by: 's avatarTobias Koppers <1365881+sokra@users.noreply.github.com>
      08aab14b
    • next-js-bot[bot]'s avatar
      v16.4.0-canary.46 · 725705bd
      next-js-bot[bot] authored
      725705bd
    • Sebastian "Sebbie" Silbermann's avatar
      Revert "Add experimental custom webpack support" (#99227) · 58ad0793
      Sebastian "Sebbie" Silbermann authored
      This produces a new dependency chain which can lead to duplicate Next.js
      version in monorepo setups.
      
      Specifically, the internal sync is broken. It can probably be fixed with
      some dependency graph doctoring but we need to investigate and document
      that first. We're close to a stable release and the sync has been
      blocked all week so getting it back on track has priority.
      
      Reverts https://github.com/vercel/next.js/pull/98862
      58ad0793
    • Tobias Koppers's avatar
      test: stabilize lazy dynamic import error overlay (#98730) · 3192e47e
      Tobias Koppers authored
      ### What?
      
      Stabilizes the Turbopack development test that verifies an invalid
      dynamic-import target stays unparsed until activation, while checking
      both server compilation output and the browser-visible error overlay
      afterward.
      
      ### Why?
      
      Activating the target can trigger an incidental hard reload as Fast
      Refresh recovers from the error. A one-shot overlay source read can race
      that document replacement. The test should check the error it owns
      without requiring a reload or accepting an absent overlay.
      
      ### How?
      
      The test retains canary's CLI-output assertion and independently retries
      a live browser-redbox source assertion filtered to the expected
      parse-error marker. Playwright re-resolves the locator across document
      changes; both signals must actually contain the marker.
      
      ### Verification
      
      - Focused Turbopack dev case under six competing CPU workers: 20/20
      passed on rebased canary
      - Complete `lazy-dynamic-imports` Turbopack dev file: 3/3 passed on
      rebased canary
      - Related `next/dynamic` error-overlay case: 10/10 passed on rebased
      canary
      - Prettier, ESLint, clean merge-tree, and diff checks
      
      <!-- NEXT_JS_LLM -->
      
      <!-- fleet 2dd4f041-bfb3-4572-9b57-5cb45a2410cd -->
      
      ---------
      Co-authored-by: 's avatarvercel-fleet-prod[bot] <318278635+vercel-fleet-prod[bot]@users.noreply.github.com>
      Co-authored-by: 's avatarTobias Koppers <1365881+sokra@users.noreply.github.com>
      3192e47e
    • Luke Sandberg's avatar
      turbo-tasks-gc: tear down reverse edges when collecting a task (#99145) · fea76de6
      Luke Sandberg authored
      When deleting a task, remove the reverse dependeny edges also. Generally
      this should be empty, but there can be tasks if those tasks are owned by
      a root that is pinned but not active. So in the rare case that such
      tasks are 'reactivated' this also dirties the task.
      
      This closes a hole where a task with a cell_dependency can reference a
      deleted task. Then new test demonstrates the bug, without the fix it
      will panic when collecting the second task
      fea76de6
    • Tobias Koppers's avatar
      Sequence CI for branch-based PR stacks (#99086) · 10c5a6ac
      Tobias Koppers authored
      ### What?
      
      Replace the Graphite-specific CI optimizer with a **read-only TypeScript
      gate** for PR stacks formed by ordinary GitHub base/head branch links.
      This PR includes the work previously reviewed in #99116 and #99122.
      
      ### Why?
      
      Rebasing a large stack starts expensive `build_and_test` jobs on every
      PR at once. Let the first three PRs and the top PR run immediately, but
      defer middle PRs until nearby CI provides a useful signal. A deferred PR
      must **not** appear mergeable just because its jobs have not started.
      
      ### How?
      
      - **Find the stack from branch relationships**, not GitHub stack
      metadata. Middle PRs poll their three nearest predecessors every **five
      minutes**. Any predecessor with a successful, current-head/current-base
      `thank you, next` check releases full CI; three unsuccessful results
      **fail the gate and required aggregate**. While waiting, the gate stays
      in progress. After five hours without a decision, it releases full CI
      rather than stranding the PR.
      - **Revalidate before deciding.** Refresh PR and check data before
      release/failure so rebases, retargeting and reruns cannot make a stale
      check decisive. Transient GitHub 5xx/429 errors retry; permanent errors
      or ambiguous topology start full CI rather than claiming CI passed.
      - **Keep polling cheap and read-only.** Reuse PR head/base data from
      `pulls.list` (11 → **8 REST reads** per ordinary waiting poll), while
      still checking *all three* predecessors. The bundled action is checked
      out at the current test-merge `${{ github.sha }}` without persisted
      credentials; forks bypass the checkout/gate and run full CI. The
      six-hour gate stays on `ubuntu-latest` because `ubuntu-slim` has a
      15-minute limit.
      
      <details>
      <summary>Implementation and review tradeoffs</summary>
      
      - The local action bundles a pinned Octokit. Isolated Jest tests use
      fake timers and a real check-run response fixture. CI runs the existing
      Next.js lint/examples/externals/browser checks before action-specific
      install, typecheck, build and Jest; a generated-diff guard catches stale
      bundles.
      - `uses: ./...` requires action files already in the job workspace; it
      does not fetch them like `owner/repo@ref`. The sparse checkout fetches
      only `action.yml` and `dist/index.js`. Caller and reusable workflow
      grant read-only `checks`, `contents` and `pull-requests` permissions;
      the gate inherits no secrets.
      - Cheap-check graph splitting was deliberately deferred. Job-level
      concurrency does not enforce predecessor *success*; stopping at the
      first pending predecessor misses an older success. GraphQL batching
      needs a verified check-association projection and measured cost before
      replacing the current REST lookup.
      - When the child commits became reachable from this branch, GitHub
      automatically marked #99116 and #99122 merged; no PR merge operation was
      used. [Action
      README](https://github.com/vercel/next.js/blob/ci/branch-stack-polling-gate/.github/actions/pr-stack-ci-gate/README.md)
      documents the contract.
      
      </details>
      
      ### Verification
      
      - **Focused checks:** TypeScript, Prettier, deterministic bundle/diff
      checks and **24/24 Jest** passed. The final comment-only change in
      `src/gate.ts` produced a byte-identical bundle. The [current root
      run](https://github.com/vercel/next.js/actions/runs/35967795126) passed
      its gate, lint and required `thank you, next` on **attempt 2** (same
      SHA); attempt 1 failed an intermittent redbox test.
      - **Live six-PR test on the compiled action:** the first three and top
      passed the gate immediately. Middle #99097 and #99098 **waited**. On
      initial attempts, both **failed closed** when their three predecessors
      were unsuccessful, with no expensive jobs started. On rerun, #99097
      **released on #99086's success** despite #99096 failing and #99095 still
      pending; full CI ran but failed unrelated tests. #99098 waited for *its
      own* three predecessors, then failed closed again. Gate release never
      marked full CI successful.
      - **Known limitation:** product-test failures—not the gate—left four
      immediate/released PRs with failed full CI on various attempts. All five
      test drafts remain open for reuse. Workflow reruns in this experiment
      were initiated externally; this agent's GitHub integration received 403
      (`actions=write` required) when attempting a rerun.
      
      <details>
      <summary>Current compiled-action six-PR results, with workflow and gate
      logs (2026-09-24)</summary>
      
      Each draft is one Jest assertion commit over its signed parent; the
      cumulative focused suite passes **24/24**. Times below are UTC.
      “Required” refers to `thank you, next`, not the gate check.
      
      | Position | PR / signed head / workflow | Gate outcome | Required check
      | Expensive work |
      | :-- | :-- | :-- | :-- | :-- |
      | 1 |
      [#99086](https://github.com/vercel/next.js/actions/runs/35967795126)
      `85444c9e` | immediate pass | failed attempt 1; **passed attempt 2** |
      ran twice |
      | 2 |
      [#99095](https://github.com/vercel/next.js/actions/runs/35969902709)
      `bc19787a` | immediate pass | failed attempts 1 and 2 | ran twice |
      | 3 |
      [#99096](https://github.com/vercel/next.js/actions/runs/35970016367)
      `98bfab58` | immediate pass | failed | ran |
      | 4 |
      [#99097](https://github.com/vercel/next.js/actions/runs/35970109573)
      `e260bf96` | [attempt
      1](https://github.com/vercel/next.js/actions/runs/35970109573/job/107537584401):
      waited → failed 08:02; [attempt
      2](https://github.com/vercel/next.js/actions/runs/35970109573/job/107609256427):
      **released 11:19** | failed both attempts (attempt 2: product tests) |
      none on attempt 1; full CI ran on attempt 2 |
      | 5 |
      [#99098](https://github.com/vercel/next.js/actions/runs/35970206069)
      `01222d4a` | [attempt
      1](https://github.com/vercel/next.js/actions/runs/35970206069/job/107537887530):
      waited → failed 08:03; [attempt
      2](https://github.com/vercel/next.js/actions/runs/35970206069/job/107609367420):
      waited → failed 11:50 | failed both attempts | **never started** |
      | 6 (top) |
      [#99099](https://github.com/vercel/next.js/actions/runs/35970312588)
      `c6f3cf4a` | immediate pass | failed attempts 1 and 2 | ran twice |
      
      **Decision evidence:** On attempt 1, #99097 saw #99096/#99095/#99086 all
      unsuccessful, and #99098 saw #99097/#99096/#99095 all unsuccessful;
      their required checks failed and downstream jobs were skipped *because
      their gates failed*. After #99086's successful rerun, #99097 attempt 2
      logged `#99096=unsuccessful, #99095=waiting, #99086=success`; its gate
      passed and [build-next actually ran and
      passed](https://github.com/vercel/next.js/actions/runs/35970109573/job/107609336752).
      #99098 could not use #99086 (outside its nearest-three window); it
      failed at 11:50 when #99097/#99096/#99095 had all finished without
      success. Completed failed gates do not restart automatically when a
      predecessor reruns.
      
      </details>
      
      <details>
      <summary>Unrelated CI failures and recovered flaky tests</summary>
      
      - #99086's attempt 1 failed `lazy-dynamic-imports › does not parse a
      dynamic import target before activation`: redbox source was `null` after
      all in-job retries
      ([job](https://github.com/vercel/next.js/actions/runs/35967795126/job/107530613870)).
      **The same-head workflow passed on attempt 2.**
      - #99095, #99096, #99097 attempt 2 and #99099 failed
      `instant-insights-tab-overlay › should wrap the mobile overlay header
      only when it does not fit`: observed header top difference **24**,
      expected **< 4**, after in-job retries ([example
      job](https://github.com/vercel/next.js/actions/runs/35969902709/job/107537341995)).
      #99096 also failed `turbopack-loader-file-dependencies › should update
      when a build dependency changes` (`build-one` instead of `build-two`,
      [job](https://github.com/vercel/next.js/actions/runs/35970016367/job/107537932737)).
      #99097 attempt 2 additionally failed the `lazy-dynamic-imports` redbox
      assertion in an experimental cache-components shard
      ([job](https://github.com/vercel/next.js/actions/runs/35970109573/job/107609732882)).
      These failures are not gate-test failures.
      - **Recovered in-job flakes:** #99096
      `instant-validation/head-and-reporting › invalid - runtime prefetch -
      dynamic viewport blocks navigation` (redbox did not open; passed retry
      1/2); #99095 `use-cache-without-experimental-flag › should recover from
      the build error if useCache flag is set` (Playwright execution context
      destroyed on navigation; passed retry 1/2); #99097 attempt 2
      `lazy-dynamic-imports` in a different Turbopack-dev shard (passed retry
      2/2), `instant-validation/suspense-boundaries` (redbox did not open;
      passed retry 2/2), and
      `instant-validation/head-and-reporting.partial-prefetching` (redbox did
      not open; passed retry 1/2).
      
      </details>
      
      <details>
      <summary>Earlier implementation smoke tests and CJS-gate stack
      runs</summary>
      
      - The [folded-code root head `89f0ca34` full
      run](https://github.com/vercel/next.js/actions/runs/35965443069) passed
      gate, build, lint (including **24/24 Jest**) and required `thank you,
      next`. The earlier [compiled-action child head `25647f00`
      run](https://github.com/vercel/next.js/actions/runs/35922700289) also
      passed full CI, proving same-SHA sparse checkout before the lint-order
      follow-up was folded into root.
      - **CJS-gate fail-closed run (base `ec0abf41`, 2026-09-23):**
      [#99086](https://github.com/vercel/next.js/actions/runs/35875541530),
      [#99095](https://github.com/vercel/next.js/actions/runs/35875598833),
      [#99096](https://github.com/vercel/next.js/actions/runs/35875651890) and
      top [#99099](https://github.com/vercel/next.js/actions/runs/35875952426)
      ran full CI and failed unrelated tests. Middle
      [#99097](https://github.com/vercel/next.js/actions/runs/35875701404/job/107231099534)
      failed its gate at 15:20 and
      [#99098](https://github.com/vercel/next.js/actions/runs/35875873698/job/107231827956)
      at 15:21; required checks failed and expensive jobs never started. A
      Turbopack source-map snapshot failure was subsequently fixed in
      [#99106](https://github.com/vercel/next.js/pull/99106).
      - **CJS-gate release run (base `8ff7f2bd`, 2026-09-23):**
      [#99086](https://github.com/vercel/next.js/actions/runs/35887913444) and
      [#99095](https://github.com/vercel/next.js/actions/runs/35887957948)
      passed full CI.
      [#99097](https://github.com/vercel/next.js/actions/runs/35888074168/job/107273254039)
      and
      [#99098](https://github.com/vercel/next.js/actions/runs/35888146874/job/107273514756)
      waited, then released at the next poll on #99095's green check while a
      nearer predecessor was unresolved. Their later full CI failed; top
      [#99099](https://github.com/vercel/next.js/actions/runs/35888202575)
      started immediately (later canceled). These precede the TypeScript
      action and are not claimed as evidence for its current head.
      
      </details>
      
      <!-- NEXT_JS_LLM -->
      
      
      <!-- fleet d2ae64f1-7106-4ba4-b88a-7785c19904c6 -->
      
      ---------
      Co-authored-by: 's avatarvercel-fleet-prod[bot] <318278635+vercel-fleet-prod[bot]@users.noreply.github.com>
      Co-authored-by: 's avatarTobias Koppers <1365881+sokra@users.noreply.github.com>
      10c5a6ac
    • Andrew Clark's avatar
      Prefetch scheduler mirrors the navigation's tree walk (#98975) · 34fdbdcc
      Andrew Clark authored
      The prefetch scheduler's static walk decided at the parent which child
      was still part of the current page and which began the new part of the
      route, and it fetched the head through two functions of its own that
      copied the per-segment decision (a static attempt, then a runtime deopt
      if that wasn't enough) for a node that hangs off the route root instead
      of sitting in the tree.
      
      Give the walk the same shape as the navigation's. The walk over the part
      of the route the current page also has owns the comparison for its own
      node — first whether the route position still matches, then whether any
      of its params changed — and hands off to the walk over the new part of
      the route when either fails, so the root enters the shared walk like
      every other node. The per-segment decision, a static attempt or a
      runtime deopt, lives in one function that the new-part walk calls for
      every segment, and the runtime request's walk gets the same split
      between iterating the children and deciding for one node.
      
      The head is a one-node route tree like any other segment now, so it goes
      through the same functions: the new-part static walk and the direct
      runtime fetch it already had.
      
      This is a pure restructuring and issues the same requests as before: a
      segment the navigation keeps is still prefetched at the ordinary static
      tier without a runtime deopt, the head is still fetched whenever
      anything is, and a chain of inlined segments is still dropped where a
      Shell-phase walk crosses from a kept node into the new part. A later
      change uses the shape to prefetch only what the navigation would fetch.
      34fdbdcc
    • Andrew Clark's avatar
      Separate route structure comparison from param comparison (#98974) · b3512c2e
      Andrew Clark authored
      Changing a page’s search params doesn’t change which page it is, but it
      may change the data that page needs. The router currently mixes these
      concerns by appending search params to `__PAGE__` segments. Remove that
      encoding so we can compare which route is being rendered separately from
      the param values used to render it.
      
      Param values are now compared through the existing VaryPath data
      structure, while route structure is compared during the existing tree
      traversals. Each traversal can decide which comparisons matter for the
      work it’s doing.
      
      Client-side comparisons and history restoration still need the search
      params, so store them in a separate FlightRouterState slot for now. As
      before, previous-page search params are stripped from normal request
      headers. This is a temporary step that lets us migrate incrementally.
      Eventually, FlightRouterState will be replaced by a type that represents
      the route and its params more directly.
      
      This is mostly a refactor, though it fixes an accidental inconsistency
      in search-param handling compared with regular route params, avoiding
      some redundant prefetch work.
      
      Fully cached pages already track their vary params correctly. This
      separation prepares us to do the same for partially dynamic segments
      during navigation, so changing an unrelated param won’t require
      rendering their dynamic content again.
      b3512c2e
    • Andrew Clark's avatar
      Store the head as its own CacheNode (#98973) · aa8e252f
      Andrew Clark authored
      Throughout most of the client router implementation, the head is treated
      as if it belongs to the page segment. For example, CacheNode has
      separate `rsc` and `headRsc` fields, and the page's CacheNode owns both.
      This had led to a bunch of special cases related to the head.
      
      The Segment Cache largely (though not entirely) avoids this by instead
      modeling the head as a special kind of segment. It has its own vary
      path, its own SegmentCacheEntry, etc. There are still a few places where
      treatment of the head is special, because it doesn't exist inside the
      normal route tree structure. But we treat it like segment wherever
      possible.
      
      This PR updates the navigation implementation to use the same model:
      treat the head as a segment that is a sibling to the route segment tree.
      
      This is a large diff but the changes are almost entirely mechanical.
      Although the added/removed balance is about even, that includes a new
      test case. The net number of implementation lines has gone down.
      aa8e252f
    • Andrew Clark's avatar
      Record which params a CacheNode's data depends on (#98972) · 2f19ae8e
      Andrew Clark authored
      Adds a `varyParams` field to CacheNode. When present, they represent the
      set of params that the data for that segment varies by. (When null, the
      set is unknown.)
      
      Currently, the vary params are only known when the data is fully cached;
      vary params are not recorded during a dynamic render. However, this will
      change in a future PR once we've integrated with the experimental
      Ledgers feature in React.
      2f19ae8e
    • Andrew Clark's avatar
      Use RouteTree<T> for CacheNode tree (#98971) · 9937a54a
      Andrew Clark authored
      Refactors the CacheNode tree to use the RouteTree<T> data type.
      CacheNode used to be a tree-shaped object, with a similar structure to
      RouteTree<T>. CacheNode is now a data container instead, embedded inside
      a RouteTree structure.
      
      So the usage sites change from CacheNode to RouteTree<CacheNode>.
      
      As a result, we can now diff against the render tree directly during
      prefetches and navigations, instead of diffing against
      FlightRouterState. This gets us closer to migrating everything away from
      FlightRouterState to our shared data types.
      9937a54a
    • Andrew Clark's avatar
      Remove the page/layout distinction from RouteTree (#98970) · 03383c3d
      Andrew Clark authored
      Part of a series of internal refactors to improve param tracking on the
      client.
      
      A route tree node used to come in two variants, a layout and a page,
      distinguished by an isPage flag that every constructor had to keep
      consistent with the VaryPath it built. The variants only differed in
      whether the vary path carried a search params entry, and the code that
      needed to know either checked the flag or read the vary path's shape
      positionally.
      
      Collapse them into one RouteTree type with one VaryPath type. The few
      places that still care whether a node is a page derive it from the
      segment. When something needs to read the search params, they can be
      accessed from the RouteTree or VaryPath like we do for other params.
      
      This is a pure refactor of the existing code. It reads more naturally
      with a single tree type and a single vary path type, which the rest of
      this stack builds on.
      03383c3d
    • Luke Sandberg's avatar
      Fix root predicates (#98942) · cb95ca37
      Luke Sandberg authored
      Fix the definition of GC roots to account for all transient references
      
      There was a categorical mistake in how `uppers` and
      `collectible_dependents` were interpreted. Both of these refer to a task
      above us and this is it is a reference but the reference might be
      comming from transient or persistent tasks. This is what kept confusing
      the 'root' definition
      
      So the new definition is simply
      
      * You are collectible if there are no references (persistent or
      transient)
      * You are a root if there are only transient references, which includes
      transient uppers and collectible_dependents
      
      
      The old definition expected roots to have no uppers at all, which isn't
      correct. A `run_once` task becomes an aggregation root and so the nested
      persistent tasks reference it as an upper. Those tasks should become gc
      roots since they are on the transient <--> persistent boundary and the
      only thing retaining them is references from transient tasks.
      cb95ca37
    • Luke Sandberg's avatar
      [turbopack] only remove followers if they are actually removed as children (#98843) · c947937b
      Luke Sandberg authored
      ## What
      
      When removing edges, only enqueue `lost follower` jobs when we actually
      remove children.
      
      ## Why
      
      GC can race with other kinds of task completion which also remove
      outgoing edges.
      
      If task completion drops a child from a task that also becomes
      collectible. Then it is possible for there to be an enqueued
      CleanupOldEdges job at the time GC runs. Then we will have two jobs that
      remove children, only one will succeed but both will enqueue
      `InnerOfUppersLostFollowersJob` for all the chidren. This can lead to
      panics since one of those jobs will fail to remove the follower
      
      The fix is simple, only remove tasks as followers when you actually
      removed them as children.
      c947937b
    • Luke Sandberg's avatar
      turbo-tasks-backend: simplify the collectibility predicate and collect transient tasks (#98615) · 645cae03
      Luke Sandberg authored
      Fix refcounting for transient tasks
      - ensure transient tasks refcount each other correctly, and are eligible
      for collection
      - ensure root tasks are born with a transient_ref_count of 1
      
      Also drop some ultimately harmful conditions from the gc predicate
      - don't consider `cell_dependent`. cell dependencies form cycles and are
      subsumed by ancestors
      - to read a cell you have to read it from a child (covered by
      parent_count) or be passed it from a parent. For the parent to have
      passed it to you it must have a child dependency on the producer or be
      passed it by its parent (recursively). So the task that passed you the
      vc must be holding ownership over both tasks.
      - therefore a stale cell_dependency can only exist within a task that is
      effectively dead, however because we do not reliably clean up all tasks
      in a session (due to leaking root tasks), we cannot fully exclude the
      existence of stale cell_dependency edges. We just know they must only
      exist within unreachable tasks.
      
      ---------
      Co-authored-by: 's avatarClaude Opus 5 (1M context) <noreply@anthropic.com>
      645cae03
    • Luke Sandberg's avatar
      turbo-tasks-gc: Invalidating a deleted task should be a no-op (#98614) · cc7e804c
      Luke Sandberg authored
      Make invalidating deleted tasks a no-op.
      
      Invalidators now represent a 'weak' reference to a task, and
      invalidations don't assert task existence. We know that the task exists
      when the invalidator is created but by the time an invalidation occurs
      it might not exist anymore. So treat the dependency as weak.
      
      Take care to not create 'blank' tasks when querying for them, this can
      confuse assertions that occur later.
      
      
      ## Why is this safe?
      
      Currently it is safe because we never reuse task ids, so invalidating a
      deleted task is perfectly reasonable, much like invalidating a
      non-active task, there is nothing to do
      
      In the future if we start reusing task ids, then this becomes an ABA
      problem. An invalidator can point at a new task. This is also not too
      bad, a spurious invalidation is self healing. Also most invalidators are
      not actually persisted which limits the risk.
      
      So we can either decide to build a mechanism to tear down these stale
      edges or tolerate spurious invalidations. I think we could make
      invalidator users subscribe to a 'deleted task id bus' which would allow
      them to drop references, this could be useful and then the few cases
      where we _persist_ invalidators (by way of `State` objects) would need
      to devise a new mechanism (or decide to tolerate the spurious
      invalidations)
      
      ---------
      Co-authored-by: 's avatarClaude Opus 5 (1M context) <noreply@anthropic.com>
      Co-authored-by: 's avatarvercel[bot] <35613825+vercel[bot]@users.noreply.github.com>
      cc7e804c
    • Josh Story's avatar
      Avoid decoding prerender matcher parameters twice (#99121) · b9be0685
      Josh Story authored
      ## Summary
      
      Avoid decoding route parameters a second time while selecting prerender
      metadata. This fixes valid percent-containing URLs returning 500,
      including both build-generated closed routes and runtime-generated open
      routes.
      
      For example, a request for
      `/open/docs/space%20here/with%2Fslash/fallback%25` is already normalized
      for manifest lookup before reaching `PrerenderManifestMatcher`. Its
      final parameter now contains a literal `%`. The matcher previously used
      `getRouteMatcher`, which tried to decode the captures again and threw
      even though this call only needed to determine whether the route
      structure matched.
      
      Test the existing, lazily constructed route regular expression directly
      instead. Matcher precedence, source-page filtering, and the current
      server parameter encoding behavior stay unchanged. This does not
      introduce a new manifest or adapter contract.
      
      The open fallback is requested twice to cover generation and reuse. The
      existing fixture uses `dynamicParams = false` and remains excluded from
      the Cache Components matrix, which does not support that configuration.
      
      This is an independent prerequisite for #97393: the failure reproduces
      without the parameter-matching API and should be fixed for existing apps
      too.
      
      ## Verification
      
      - Before the fix, the new closed and open catch-all requests returned
      500 with a parameter-decoding error. After the fix, both return 200 and
      retain the existing encoded server parameter values.
      - The focused matcher unit suite passes all six tests.
      - The three encoding regressions pass with production Webpack after
      restacking. The API routing suite also passes all 36 tests with
      fresh-native production Turbopack above this prerequisite, including its
      existing percent-containing URL cases.
      - The earlier isolated verification passed package TypeScript and
      changed-file lint. Deployment verification remains for CI.
      
      <!-- NEXT_JS_LLM -->
      b9be0685
  2. 24 Sep, 2026 2 commits
    • next-js-bot[bot]'s avatar
      v16.4.0-canary.45 · c1ce6ea2
      next-js-bot[bot] authored
      c1ce6ea2
    • Sebastian "Sebbie" Silbermann's avatar
      [test] Fix new deploy suites (#99162) · d53f1dde
      Sebastian "Sebbie" Silbermann authored
      The `test-e2e-deploy-release` run for `v16.4.0-canary.43` failed three
      webpack shards deterministically
      ([run](https://github.com/vercel/next.js/actions/runs/35989345663))
      after #98847 and #98848 removed the deploy-mode exclusions.
      
      On PRs we only run deploy tests for Turbopack so these failures were
      missed.
      
      - `use-cache-segment-configs`: #98847 added a deploy-specific snapshot
      for Turbopack, but the webpack path still used the local snapshot.
      Vercel's fetched build logs strip code-frame indentation and trailing
      whitespace, so the deployed output can never match it. The webpack
      assertion now has its own deploy branch with the exact formatting Vercel
      produces.
      - `app-edge-invalid-reexport`: #98848 removed `skipDeployment`, but the
      fixture exports `preferredRegion = 'test-region'`, which the Vercel
      platform rejects during deployment (`Invalid region: test-region`). The
      test now overrides the basic page with a valid region (`iad1`) in deploy
      mode. The asserted re-export warnings are build-time diagnostics, so the
      region value does not affect them.
      - `deprecation-warnings`: both tests call `next.start()` inside the test
      body, which `e2e-utils` caps at 60 seconds when no explicit timeout is
      given. Deployments now take about two minutes (`Ready in 2m`, up from
      `Ready in 1m` in the passing canary.42 run with the same Vercel CLI
      59.26.0), so both tests time out on every run. They now pass an explicit
      240-second timeout, matching the other deploy tests.
      
      ---------
      Co-authored-by: 's avatarClaude Code (kimi-k3[1m]) <noreply@anthropic.com>
      d53f1dde