- 03 Nov, 2025 18 commits
-
-
nextjs-bot authored
-
Tan Zi Gang authored
Fix #85715 --------- Co-authored-by:
graphite-app[bot] <96075541+graphite-app[bot]@users.noreply.github.com> Co-authored-by:
Jiwon Choi <devjiwonchoi@gmail.com>
-
Jiwon Choi authored
Codemod `next-lint-to-eslint-cli` did not handle `extends: ['next']` in legacy ESLint config. x-ref: https://github.com/vercel/next.js/issues/85244#issuecomment-3482558412 Fixes #85244
-
ajstrongdev authored
### What? - Add mention of ESLint being the default linter when specifying `--yes` flag on `create-next-app` ### Why? - Adds consistency by calling out the default linter. Co-authored-by:Joseph <joseph.chamochumbi@vercel.com>
-
Sebastian "Sebbie" Silbermann authored
Suspecting that this fixes (or reduces) the "Watchpack Error (watcher): Error: EMFILE:" issues e.g. https://github.com/vercel/next.js/actions/runs/19028750529/job/54337767548?pr=85637 I noticed that all the Watchpack errors in CI happened in tests with a tsconfig (or ones where a tsconfig was created). When watchpack starts watching, an initial `aggregated` event is set. During that event we used to consider all files as changed because we hadn't seen them before. This also meant we considered the tsconfig as changed and invalidated the hot-reloader. This is excessive and likely hits some hard to debug race conditions. Now we check if the file was actually changed since we started watching by comparing its change time to the time when we started watching. Each file modified before we started watching is no longer considered changed.
-
Sebastian "Sebbie" Silbermann authored
Fixes https://github.com/vercel/next.js/actions/runs/18937795131/job/54068900147#step:34:1879 `--experimental-https` test creates a cert relying on no cert existing in the default path. When it passes, it doesn't cleanup. Retrying that test file due to an unrelated failure would therefore always fail. Now we ensure no cert exists.
-
Hendrik Liebau authored
[flakiness metric](https://app.datadoghq.com/ci/test/runs?query=test_level%3Atest%20%40git.repository.id%3A%22github.com%2Fvercel%2Fnext.js%22%20%40test.name%3A%22use-cache-custom-handler%20should%20use%20a%20modern%20custom%20cache%20handler%20if%20provided%22%20%40test.type%3A%22nextjs%22%20%40test.status%3A%22fail%22%20%40git.branch%3Acanary&agg_m=count&agg_m_source=base&agg_t=count&fromUser=true&index=citest&start=1759310750664&end=1761902750664&paused=false) The test page used a low revalidate time of 3 seconds. In CI, sometimes that much time passes between the first and second request, causing the test to fail when comparing the data values (locally reproducible with `{ cpuThrottleRate: 20 }`). Increasing the revalidate time to 6 seconds should make the test more robust.
-
Sebastian "Sebbie" Silbermann authored
-
Joseph authored
-
Tobias Koppers authored
### What? Makes adding multiple items at once more efficient.
-
Vercel Release Bot authored
This auto-generated PR updates font data with latest available
-
Sebastian "Sebbie" Silbermann authored
-
Jiachi Liu authored
In v16, `id` argument in metadata routes becomes promise which requries to to unwrap it first to access the value Follow up of #84491 Closes #85632 --------- Co-authored-by:Joseph Chamochumbi <joseph.chamochumbi@vercel.com>
-
Tim Neutkens authored
## What? Upgrade the image crate which includes some bugfixes for bmp files.
-
Marjan Ahmed authored
### What? Fixes an inconsistency in the v16 upgrade documentation for `generateImageMetadata`. ### Why? In the “Upgrading to Version 16” guide, the example incorrectly awaited the `params` object inside `generateImageMetadata`, implying that it was a Promise. However, according to the API reference, `params` is a normal object. This PR updates the example to reflect the correct synchronous usage. ### How? Replaced: ```ts export async function generateImageMetadata({ params }) { const { slug } = await params return [{ id: '1' }, { id: '2' }] } With ```ts export async function generateImageMetadata({ params }) { // params is a normal object return [{ id: '1' }, { id: '2' }] } --------- Co-authored-by:Joseph <joseph.chamochumbi@vercel.com>
-
Sebastian "Sebbie" Silbermann authored
-
Vercel Release Bot authored
This auto-generated PR updates the development integration test manifest used when testing Rspack.
-
Vercel Release Bot authored
This auto-generated PR updates the production integration test manifest used when testing Rspack.
-
- 02 Nov, 2025 4 commits
-
-
nextjs-bot authored
-
Vercel Release Bot authored
This auto-generated PR updates the production integration test manifest used when testing Rspack.
-
Jiwon Choi authored
The heading sections in DevTools regressed. We decided to remove them as "Dev Server" settings are obviously dev server settings, no need to differentiate as separate tabs. | Before | After | |--------|--------| | <img width="1024" height="1006" alt="CleanShot 2025-11-02 at 12 59 21@2x" src="https://github.com/user-attachments/assets/0a8b0c2f-3f0a-487d-9344-b33b5ee016f5" /> | <img width="1024" height="1006" alt="CleanShot 2025-11-02 at 12 59 07@2x" src="https://github.com/user-attachments/assets/ddd3b029-8957-4129-aaf2-7b1a57a52e00" /> |
-
Vercel Release Bot authored
This auto-generated PR updates the development integration test manifest used when testing Rspack.
-
- 01 Nov, 2025 2 commits
-
-
nextjs-bot authored
-
nextjs-bot authored
-
- 31 Oct, 2025 16 commits
-
-
nextjs-bot authored
-
Zack Tanner authored
Moves `resume-data-cache` to `test/e2e`. We don't currently run `production` tests when deployed - many of them do funky things with `next.build` / `next.start` and are meant to test local behavior rather than deployed behavior. As a result, I also temporarily disabled this test when deployed because there are currently bugs.
-
Zack Tanner authored
This removes the indirection that was set up to codesplit on the client. But now that there's only a single client cache entrypoint, we can delete the indirection and move everything to a top-level export.
-
Wyatt Johnson authored
### What? This PR preserves the specific interception route markers in dynamic parameter types throughout the routing system. ### Why? Previously, the code only tracked whether a parameter was intercepted (using generic types like `'dynamic-intercepted'` or `'catchall-intercepted'`) but not which specific interception marker was used. This meant that routing information about the interception pattern was being lost. Interception routes use different markers: - `(.)` - matches segments at the same level - `(..)` - matches segments one level up - `(..)(..)` - matches segments two levels up - `(...)` - matches segments from the root app directory Without preserving the specific marker, the routing system couldn't properly distinguish between these different interception patterns, which could lead to incorrect route matching behavior. ### How? The fix encodes the specific interception marker pattern directly in the type definition. For example: - `'dynamic-intercepted'` → `'dynamic-intercepted-(.)'`, `'dynamic-intercepted-(..)'`, etc. - `'catchall-intercepted'` → `'catchall-intercepted-(.)'`, `'catchall-intercepted-(..)'`, etc. This change is propagated throughout the routing system: - Type definitions in `app-router-types.ts` and `types.ts` - Route parameter parsing in `route-params.ts` - Static path generation in `build/static-paths/app.ts` - Dynamic parameter utilities in `get-dynamic-param.ts` and `get-segment-param.tsx` - On-demand entry handling in development mode NAR-486
-
Tobias Koppers authored
### What? add a size hint to the Chunk iterator to allow optimizations
-
Shu Ding authored
Clear global data array to release memory after server data is flushed. For large apps we found that up to megabytes of data only used by hydration can be held by `window.__next_f` for the entire session.
-
Sebastian "Sebbie" Silbermann authored
Because deploy tests won't have access to the child process.
-
Hendrik Liebau authored
This PR extracts the existing git change detection logic from the `get-changed-tests` script into a separate module. This allows reusing the same logic in the `run-for-change` script, ensuring consistent behavior across the two scripts to determine changed files. Specifically, this ensures that in stacked PRs, the docs-only change detection (and the test-only change detection in the upstack PR) works correctly by diffing against the appropriate base branch.
-
Tobias Koppers authored
### What? Turbo tasks Arc should be dropped before returning to the caller
-
Benjamin Woodruff authored
Migrated from https://github.com/vercel/next.js/pull/85478 into this non-fork because we had some issues with our CI infra that should be fixed going forward: https://vercel.slack.com/archives/C04KC8A53T7/p1761862171772939 --- fixes #85474 Previously, the analyzer treated all assignment expressions the same way, evaluating them as their right-hand side value. This was incorrect for compound assignments (+=, -=, *=, |=, >>=, etc.), which compute a new value based on both the left and right operands. This commit distinguishes between simple assignments (=) and compound assignments. Simple assignments continue to evaluate to the right-hand side, while compound assignments are now properly marked as unknown values with side effects. Also adds tests for compound assignments in conditionals, covering bitshift operations (>>=, <<=, >>>=) and bitwise operations (|=, &=, ^=) with both numbers and bigints.
-
Tobias Koppers authored
### What? Fixes a deadlock due to concurrent task locks. Also adds a debug check to prevent this in future.
-
Joseph authored
Co-authored-by:vercel[bot] <35613825+vercel[bot]@users.noreply.github.com>
-
Tobias Koppers authored
### What? Enabling the feature will double execute tasks and print to stdout when tasks are updating cells while recomputing The change will also change the behavior of cell updates during recomputation slightly. Instead of updating the cell and not notifying dependents, it will not update the cell at all ignoring the (unexpected) change. This means when a task is recomputed without being invalidated, it won't update any cell data, except for the missing cell values.
-
Tobias Koppers authored
### What? Helps finding Item does not exist or access type errors
-
Tobias Koppers authored
### What? Add a few more checks when verify_serialization is enabled
-
Sebastian "Sebbie" Silbermann authored
-