When the router cache can't find a cache node for the requested segment,
it performs a request to the server to get the missing data. This
request to the server currently will always include the `next-url`
header, and on soft-navigations, the router will route the request to
the intercepted handler. This lazy fetch is treated as a soft navigation
by the server, and will incorrectly return data for the intercepted
route.
Similar to the handling in `router.refresh`, and the server action
reducer, we should not include the `next-url` header if there's no
interception route currently in the tree, as otherwise we'll be
erroneously triggering the intercepted route.
Fixes#64676
Closes NEXT-3146
Reverts vercel/next.js#64271
This appears to be causing problems rendering error boundaries on SPA
navigations and needs further investigation
Closes NEXT-3150
### Why
If you have a client entry that mixing `default` re-export and `*`
re-export, atm we cannot statically analyze all the exports from this
the boundary, unless we can apply barrel file optimization for every
import which could slow down speed.
```js
// index.js
'use client'
export * from './client'
export { default } from './client'
```
Before that happen we high recommend you don't mixing that and try to
add the client directive to the leaf level client module. We're not able
to determine what the identifiers are imported from the wildcard import
path. This would work if we resolved the actual file but currently we
can't.
### What
When we found the mixing client entry module like that, we treat it as a
CJS client module and include all the bundle in client like before what
we have the client components import optimization.
Ideally we could warn users don't apply the client directive to these
kinda of barrel file, and only apply them to where we needed.
Fixes#64518
Closes NEXT-3119
## Info
I did not create a GitHub issue, but this PR for fixing the issue. Hope
that's okay.
## The bug
We run a 'normal' custom server described here:
https://nextjs.org/docs/pages/building-your-application/configuring/custom-server
```ts
const next = require('next');
const dev = process.env.NODE_ENV !== 'production';
const port = 3000;
const app = next({ dev, port, quiet: !dev })
const handle = app.getRequestHandler()
// ...
```
The `quiet` settings is missed in some deeper Next.js internal api
calls. For that reason, the setting did not work at all. This works only
in production mode.
## Testing
I tried to implement a test in `/test/production/custom-server` but I
failed to get the `stderr` messages. When someone give me some guidance,
I will add a new test.
Here are some criteria:
- need: custom server with `quiet: true` setting
- need: a next.js page with thrown error. e.g. `export const
getServerSideProps = () => { throw new Error('failed') }`
- need: collect all `stderr` messages
I tested the fix in our project by editing the files in `node_modules`
-- it worked.
## Background
We upgrade our project from next `13.3.4` to `14.2.0` and saw this
regression. I can not tell in which version this bug was added. We run a
Next.js + 'custom server' + Docker setup. Because we have a lot of
traffic, we need to suppress the standard Next.js logging and using our
own logging.
<!-- Thanks for opening a PR! Your contribution is much appreciated.
To make sure your PR is handled as smoothly as possible we request that
you follow the checklist sections below.
Choose the right checklist for the change(s) that you're making:
## For Contributors
### Improving Documentation
- Run `pnpm prettier-fix` to fix formatting issues before opening the
PR.
- Read the Docs Contribution Guide to ensure your contribution follows
the docs guidelines:
https://nextjs.org/docs/community/contribution-guide
### Adding or Updating Examples
- The "examples guidelines" are followed from our contributing doc
https://github.com/vercel/next.js/blob/canary/contributing/examples/adding-examples.md
- Make sure the linting passes by running `pnpm build && pnpm lint`. See
https://github.com/vercel/next.js/blob/canary/contributing/repository/linting.md
### Fixing a bug
- Related issues linked using `fixes #number`
- Tests added. See:
https://github.com/vercel/next.js/blob/canary/contributing/core/testing.md#writing-tests-for-nextjs
- Errors have a helpful link attached, see
https://github.com/vercel/next.js/blob/canary/contributing.md
### Adding a feature
- Implements an existing feature request or RFC. Make sure the feature
request has been accepted for implementation before opening a PR. (A
discussion must be opened, see
https://github.com/vercel/next.js/discussions/new?category=ideas)
- Related issues/discussions are linked using `fixes #number`
- e2e tests added
(https://github.com/vercel/next.js/blob/canary/contributing/core/testing.md#writing-tests-for-nextjs)
- Documentation added
- Telemetry added. In case of a feature if it's used or not.
- Errors have a helpful link attached, see
https://github.com/vercel/next.js/blob/canary/contributing.md
## For Maintainers
- Minimal description (aim for explaining to someone not on the team to
understand the PR)
- When linking to a Slack thread, you might want to share details of the
conclusion
- Link both the Linear (Fixes NEXT-xxx) and the GitHub issues
- Add review comments if necessary to explain to the reviewer the logic
behind a change
### What?
### Why?
### How?
Closes NEXT-
Fixes #
-->
## What?
- Changes webpack output target to `es6` (required for `async function`
output)
- Adds tests for top level await in server components and client
components (App Router)
- Converted the async-modules tests to `test/e2e`
- Has one skipped test that @gnoff is going to look into. This shouldn't
block merging this PR 👍
Adds additional tests for top level `await`.
Since [Next.js
13.4.5](https://github.com/vercel/next.js/releases/tag/v13.4.5) webpack
has top level await support enabled by default.
Similarly Turbopack supports top level await by default as well.
TLDR: You can remove `topLevelAwait: true` from the webpack
configuration.
In writing these tests I found that client components are missing some
kind of handling for top level await (async modules) so I've raised that
to @gnoff who is going to have a look.
<!-- Thanks for opening a PR! Your contribution is much appreciated.
To make sure your PR is handled as smoothly as possible we request that
you follow the checklist sections below.
Choose the right checklist for the change(s) that you're making:
## For Contributors
### Improving Documentation
- Run `pnpm prettier-fix` to fix formatting issues before opening the
PR.
- Read the Docs Contribution Guide to ensure your contribution follows
the docs guidelines:
https://nextjs.org/docs/community/contribution-guide
### Adding or Updating Examples
- The "examples guidelines" are followed from our contributing doc
https://github.com/vercel/next.js/blob/canary/contributing/examples/adding-examples.md
- Make sure the linting passes by running `pnpm build && pnpm lint`. See
https://github.com/vercel/next.js/blob/canary/contributing/repository/linting.md
### Fixing a bug
- Related issues linked using `fixes #number`
- Tests added. See:
https://github.com/vercel/next.js/blob/canary/contributing/core/testing.md#writing-tests-for-nextjs
- Errors have a helpful link attached, see
https://github.com/vercel/next.js/blob/canary/contributing.md
### Adding a feature
- Implements an existing feature request or RFC. Make sure the feature
request has been accepted for implementation before opening a PR. (A
discussion must be opened, see
https://github.com/vercel/next.js/discussions/new?category=ideas)
- Related issues/discussions are linked using `fixes #number`
- e2e tests added
(https://github.com/vercel/next.js/blob/canary/contributing/core/testing.md#writing-tests-for-nextjs)
- Documentation added
- Telemetry added. In case of a feature if it's used or not.
- Errors have a helpful link attached, see
https://github.com/vercel/next.js/blob/canary/contributing.md
## For Maintainers
- Minimal description (aim for explaining to someone not on the team to
understand the PR)
- When linking to a Slack thread, you might want to share details of the
conclusion
- Link both the Linear (Fixes NEXT-xxx) and the GitHub issues
- Add review comments if necessary to explain to the reviewer the logic
behind a change
### What?
### Why?
### How?
Closes NEXT-
Fixes #
-->
Closes NEXT-3126
Fixes https://github.com/vercel/next.js/issues/43382
### Why
For app page rendering on edge, the `AsyncLocalStorage` (ALS) should be
bundled as same instance across layers. We're accessing the ALS in
`next/dynamic` modules during SSR for preloading CSS chunks. There's a
bug that we can't get the ALS store during SSR in edge, I digged into it
and found the root cause is:
We have both import paths:
`module (rsc layer) -> request ALS (shared layer)`
`module (ssr layer) -> request ALS (shared layer)`
We expect the ALS to be the same module since we're using the same layer
but found that they're treated as different modules due to applying
another loader transform on ssr layer. They're resulted in the same
`shared` layer, but with different resource queries. This PR excluded
that transform so now they're identical across layers.
### What
For webpack, we aligned the loaders applying to the async local storage,
so that they're resolved as the same module now.
For turbopack, we leverage module transition, sort of creating a new
`app-shared` layer for these modules, and apply the transition to all
async local storage instances therefore the instances of them are only
bundled once.
To make the turbopack chanegs work, we change how the async local
storage modules defined, separate the instance into a single file and
mark it as "next-shared" layer with import:
```
any module -> async local storage --- use transition, specify "next-shared" layer ---> async local storage instance
```
Closes NEXT-3085
### What
When rendering a parallel slot multiple times in a single layout, in
conjunction with using an error boundary, the following TypeError is
thrown:
> Cannot destructure property 'parallelRouterKey' of 'param' as it is
null
### Why
I'm not 100% sure of the reason, but I believe this is because of how
React attempts to dededupe (more specifically, "detriplficate") objects
that it sees getting passed across the RSC -> client component boundary
(and an error boundary is necessarily a client component). When React
sees the same object twice, it'll create a reference to that object and
then use that reference in future places where it sees the object. My
assumption is that there's a bug somewhere here, as the `LayoutRouter`
component for the subsequent duplicated parallel slots (after the first
one) have no props, hence the TypeError.
### How
Rather than passing the error component as a prop to `LayoutRouter`,
this puts it as part of the `CacheNodeSeedData` data structure. This is
more aligned with other properties anyway (such as `loading` and `rsc`
for each segment), and seems to work around this bug as the
`initialSeedData` prop is only passed from RSC->client once.
EDIT: Confirmed this is also fixed after syncing the latest React, due
to https://github.com/facebook/react/pull/28669Fixes#58485
Closes NEXT-2095
`initialCanonicalUrl` differs from the `canonicalUrl` that gets set on
the client, such as when there's a basePath set.
This hoists the `canonicalUrl` construction up so we can re-use it when
adding refetch markers to the tree.
This also renames `pathname` -> `path` since it also includes search
params. I added a test to confirm no extra erroneous fetches happened in
both cases.
Fixes#64584
Closes NEXT-3130
Since `AppRouterState` is promise-based (so we can `use` the values and
suspend in render), the state itself isn't stable between renders. Each
action corresponds with a new Promise object. When a link is hovered, a
prefetch action is dispatched, even if the prefetch never happens (for
example, if there's already a prefetch entry in the cache, and it
doesn't need to prefetch again). In other words, the prefetch action
will be dispatched but won't necessarily change the state.
This means that these no-op actions that don't actually change the state
values will trigger a re-render. Most of the context providers in
`AppRouter` are memoized with the exception of `LayoutRouter` context.
This PR memoizes those values so that consumers are only notified of
meaningful updates.
Fixes#63159
Closes NEXT-3127
### What
When middleware.js is present, the logging is duplicated. We should
filter out the 1st middleware request and only log the actual one going
through request handler / renderer
Closes NEXT-3125
When an action is marked as "discarded", we enqueue a refresh, since the
navigation event will be invoked immediately without waiting for the
action to finish. We refresh because it's possible that the discarded
action triggered some sort of mutation/revalidation, and we want the
router to still be able to respond to that new data.
However there's a bug in this logic -- it'll only enqueue the refresh
action if there were no other actions in the queue, ignoring the case
where something is still in the queue. This makes sure that the refresh
is handled after `runRemainingActions` finishes.
When adding a test for the server component case (which doesn't hit this
refresh branch), I noticed `LayoutRouter` caused React to suspend
indefinitely, because it got stuck in the `use(unresolvedThenable)`
case. We should only suspend indefinitely if we kicked off a the
`SERVER_PATCH` action, as otherwise it's possible nothing will ever
break out of that branch.
Fixes#64517
Closes NEXT-3124
## What
Determine if the client module is a CJS file and `default` export is
imported, then we include the whole module instead of using webpack
magic comments to only extract `default` export.
## Why
Unlike ESM, The `default` export of CJS module is not just `.default`
property, we need to include `__esModule` mark along with `default`
export to make it function properly with React client module proxy
Fixes#64518
Closes NEXT-3119
### What
This PR's goal is to improve the throughput performance of the Next.js
server when handling Pages Router routes.
Note that the results from this are very synthetic and do not represent
the real-life performance of an application. If we only wanted to handle
hello worlds, we could probably make this even faster but on production,
a slow fetch call to your DB is probably what's slowing you down.
I'll look into App Router next.
### Why?
I guess I got nerd-sniped into it 😃
### How?
A few optimizations:
- I looked deeply at the pipeline for rendering a Pages Router page. I
noticed a lot of intermediary streams being created here and there to
eventually be concatenated to a simple string. I think this is probably
left over code from when we wanted to support streaming there and so
there's some code that was shared with the App Router, which we
absolutely don't need I think. I refactored it to be slightly simpler
with just a few string concats here and there.
- misc: I removed some redundant Promises being created here and there
and added a small inline optimisation to eliminate `if (renderOpts.dev)`
code in production.
### Nummies
Test setup: hello world pages router app, next start + autocannon
- requests handled in 10s: 18k -> 33K, **~80% improvement**
- avg latency: 4.89ms -> 2.8ms, **~42% improvement**
- avg req/res: 1846.5 -> 2983.5, **~61% improvement**
Before
<img width="742" alt="image"
src="https://github.com/vercel/next.js/assets/11064311/658e7ade-eba7-4604-a7c9-619bd51a5ec8">
vs
after
<img width="880" alt="image"
src="https://github.com/vercel/next.js/assets/11064311/2f46cf69-b788-4db2-bf90-6f65dc7abd82">
<!-- Thanks for opening a PR! Your contribution is much appreciated.
To make sure your PR is handled as smoothly as possible we request that
you follow the checklist sections below.
Choose the right checklist for the change(s) that you're making:
## For Contributors
### Improving Documentation
- Run `pnpm prettier-fix` to fix formatting issues before opening the
PR.
- Read the Docs Contribution Guide to ensure your contribution follows
the docs guidelines:
https://nextjs.org/docs/community/contribution-guide
### Adding or Updating Examples
- The "examples guidelines" are followed from our contributing doc
https://github.com/vercel/next.js/blob/canary/contributing/examples/adding-examples.md
- Make sure the linting passes by running `pnpm build && pnpm lint`. See
https://github.com/vercel/next.js/blob/canary/contributing/repository/linting.md
### Fixing a bug
- Related issues linked using `fixes #number`
- Tests added. See:
https://github.com/vercel/next.js/blob/canary/contributing/core/testing.md#writing-tests-for-nextjs
- Errors have a helpful link attached, see
https://github.com/vercel/next.js/blob/canary/contributing.md
### Adding a feature
- Implements an existing feature request or RFC. Make sure the feature
request has been accepted for implementation before opening a PR. (A
discussion must be opened, see
https://github.com/vercel/next.js/discussions/new?category=ideas)
- Related issues/discussions are linked using `fixes #number`
- e2e tests added
(https://github.com/vercel/next.js/blob/canary/contributing/core/testing.md#writing-tests-for-nextjs)
- Documentation added
- Telemetry added. In case of a feature if it's used or not.
- Errors have a helpful link attached, see
https://github.com/vercel/next.js/blob/canary/contributing.md
## For Maintainers
- Minimal description (aim for explaining to someone not on the team to
understand the PR)
- When linking to a Slack thread, you might want to share details of the
conclusion
- Link both the Linear (Fixes NEXT-xxx) and the GitHub issues
- Add review comments if necessary to explain to the reviewer the logic
behind a change
### What?
### Why?
### How?
Closes NEXT-
Fixes #
-->
Closes NEXT-3103
### Summary
Fixes#52935
`next/script` has a `Script` component that supports an `async` prop.
However, when scripts are loaded with the `async` prop set to false, the
script is loaded as if async was set to true. This may cause scripts to
execute out of order. Repro:
https://github.com/domdomegg/next-async-script-reproduction
I think this is occurring because Next uses setAttribute to set the
async and defer attributes. However, this is not a valid way to set
these properties on a script. This is because . Demo:
https://jsfiddle.net/6ktpfae1/
This PR fixes this behaviour by using removeAttribute after calling
setAttribute (rather than using setAttribute "false"). This appears to
result in correct behaviour.
Given it appears this workaround was already applied in `next/head`,
I've harmonised the code between these two.
### Next steps
I think this PR is ready for review. I acknowledge there are no test
changes, but there are no existing tests for `next/script` at all and
creating them I think would be disproportionally difficult given issues
in #52943.
---------
Co-authored-by: Tim Neutkens <tim@timneutkens.nl>
Co-authored-by: Sam Ko <sam@vercel.com>
Co-authored-by: JJ Kasper <jj@jjsweb.site>
Resolves#64412
This adds a client transition to the app route `ModuleAssetContext` and
the corresponding transforms so that client components can be safely
imported and referenced (as their proxies) in app routes.
Test Plan: Added an integration test
Closes PACK-2964
cc @shuding
CI run failed after #63053 on turbo production, which had not yet been
targeted for turbo.
x-ref: #63103
---------
Co-authored-by: JJ Kasper <22380829+ijjk@users.noreply.github.com>
Co-authored-by: Leah <8845940+ForsakenHarmony@users.noreply.github.com>
Co-authored-by: Jiachi Liu <inbox@huozhi.im>
### What?
The string value of `Metadata.openGraph.emails` throws error:
```tsx
export const metadata = {
openGraph: {
type: 'article',
emails: 'author@vercel.com',
},
};
```
Error:
```sh
r.map is not a function
```
The type is:
```tsx
type OpenGraphMetadata = {
// ...
emails?: string | Array<string>
}
```
### Why?
Basic values such as `emails` and `phoneNumbers` were not included when
`ogType` was set while resolving the values.
### How?
Include basic values when resolving the metadata.
Fixes#63415
This ensures we properly skip calling a fetch during build-time that has
`cache: 'no-store'` as it should only be called during runtime instead.
Fixes: https://github.com/vercel/next.js/issues/64462
Closes NEXT-3114
## Why?
For multi-root layouts (route groups on the root with their root
layouts, no layout file on the root), it is not possible to use global
`not-found` since the `layout` is missing on the root.
```sh
.
└── app/
├── (main)/
│ └── layout.js
├── (sub)/
│ └── layout.js
└── not-found.js --> ERR: missing layout
```
Current Behavior:
```sh
not-found.js doesn't have a root layout. To fix this error, make sure every page has a root layout.
```
## What?
Let multi-root layouts also benefit from the global `not-found`.
## How?
Wrap root `not-found` with default layout if root layout does not exist.
Although this solution is not `multi-root` specific, it won't produce
critical issues since a root `layout` is required for other cases.
Fixes#55191#54980#59180
### 🤔 What's in there?
We've deprecated config's `analyticsId` in 14.1.1 [almost 3 months
ago](https://github.com/vercel/next.js/releases/tag/v14.1.1-canary.2).
Users can opt in fot `@vercel/speed-insights`, or use
`useReportWebVitals` to report to any provider they'd like.
This PR:
- removes `analyticsId` key from configuration
- stops setting `__NEXT_PUBLIC_ANALYTICS_ID` env variable when the key
was present
- stops injecting `performance-relayer` file, when the variable is set
- cleans up related test code.
### What
Inherit the client boudaries from the found matched target in load
barrel
### Why
The root cause with the barrel transform, we missed the client boundary
directive during the transform.
Since the new version of mui's case looks like this:
Import path
page.js (server module) -> `<module>/index.js` (shared module) ->
`<module/subpath>/index.js` (client module) ->
`<module/subpath/sub-module.js> (client module)
After our transform, we lost the `"use client"` which causes the
mismatch of the transform:
In `rsc` layer: the file pointing to the sub module entry
(`<module/subpath>/index.js`), but without the client boundary.
Fixes#64369
Closes NEXT-3101
Since the router processes events sequentially, we special case
`ACTION_NAVIGATE` events to "discard" a pending server action so that it
can process the navigation with higher priority. We should apply this
same logic to `ACTION_RESTORE` events (e.g. `router.back()`) for the
same reason.
Fixes#64432
Closes NEXT-3098
`fetchCache` is a more fine-grained segment level cache-control
configuration that most people shouldn't have to use. Current semantics
of `dynamic = "force-dynamic"` will still treat the fetch as cacheable
unless explicitly overriding it in the fetch, or setting a segment level
`fetchCache`.
The `dynamic` cache configuration should be seen as a "top-level"
configuration, while more fine-grained controls should inherit logical
defaults from the top-level. Otherwise this forces people to opt-into
the `fetchCache` configuration, or manually override each `fetch` call,
which isn't what you'd expect when forcing a segment to be dynamic.
This will default to not attempting to cache the fetch when
`force-dynamic` is used. As a result, I had to update one of the
`app-static` tests to use `revalidate: 0` rather than `force-dynamic`,
as the revalidate behavior is slightly different in that it won't modify
the revalidation time on a fetch if it's non-zero.
Closes NEXT-2067
The fix is not correct, we already have request ALS wrapping the request
in app-render, shouldn't get another one for edge adapter. This is
actually a bundler bug
Reverts vercel/next.js#64382
Closes NEXT-3097
### What
Wrap async local storage for all edge runtime routes in adapter
Basically fixed the case reported in [this
tweet](https://x.com/keegandonley/status/1778538456458854880)
### Why
We're relying on the ALS for dynamic css preloading but we didn't wrap
the ALS for request handlers for edge. So if you have CSS imports in
`next/dynamic` in edge runtime it would break.
Closes NEXT-3085
<!-- Thanks for opening a PR! Your contribution is much appreciated.
To make sure your PR is handled as smoothly as possible we request that
you follow the checklist sections below.
Choose the right checklist for the change(s) that you're making:
## For Contributors
### Improving Documentation
- Run `pnpm prettier-fix` to fix formatting issues before opening the
PR.
- Read the Docs Contribution Guide to ensure your contribution follows
the docs guidelines:
https://nextjs.org/docs/community/contribution-guide
### Adding or Updating Examples
- The "examples guidelines" are followed from our contributing doc
https://github.com/vercel/next.js/blob/canary/contributing/examples/adding-examples.md
- Make sure the linting passes by running `pnpm build && pnpm lint`. See
https://github.com/vercel/next.js/blob/canary/contributing/repository/linting.md
### Fixing a bug
- Related issues linked using `fixes #number`
- Tests added. See:
https://github.com/vercel/next.js/blob/canary/contributing/core/testing.md#writing-tests-for-nextjs
- Errors have a helpful link attached, see
https://github.com/vercel/next.js/blob/canary/contributing.md
### Adding a feature
- Implements an existing feature request or RFC. Make sure the feature
request has been accepted for implementation before opening a PR. (A
discussion must be opened, see
https://github.com/vercel/next.js/discussions/new?category=ideas)
- Related issues/discussions are linked using `fixes #number`
- e2e tests added
(https://github.com/vercel/next.js/blob/canary/contributing/core/testing.md#writing-tests-for-nextjs)
- Documentation added
- Telemetry added. In case of a feature if it's used or not.
- Errors have a helpful link attached, see
https://github.com/vercel/next.js/blob/canary/contributing.md
## For Maintainers
- Minimal description (aim for explaining to someone not on the team to
understand the PR)
- When linking to a Slack thread, you might want to share details of the
conclusion
- Link both the Linear (Fixes NEXT-xxx) and the GitHub issues
- Add review comments if necessary to explain to the reviewer the logic
behind a change
### What?
### Why?
### How?
Closes NEXT-
Fixes #
-->
### What?
This adds a missing test case for a page which accesses dynamic API's
within `generateMetadata`. It also adds a test case to verify that any
of the URL's being tested can succeed based on the components link
components (it relies on finding a link to click in the browser).
Closes NEXT-3004
## Why
Currently, Server Action handlers are just normal routes and they
accepting POST requests. The only way to differentiate a normal request
(page access, API route, etc.) from a Server Action request is to check
that if it's a POST and has a Server Action ID set via headers or body
(e.g. `multipart/form-data`).
Usually, for existing page and API routes the correct handlers (page
renderer, API route handler) will take precedence over Server Action's.
But if the route doesn't exist (e.g. 404) it will still go through
Server Action's handler and result in an error. And we're eagerly
logging out that error because it might be an application failure, like
passing a wrong Action ID.
## How
In this PR we are making sure that the error is only logged if the
Action ID isn't `null`. This means that it's an intentional Server
Action request with a wrong ID. If the ID is `null`, we just handle it
like 404 and log nothing.
Fixes#64214.
Closes NEXT-3071
### What
Bump the edge runtime manifest version and add `environments` property
to each route for inlining values for deployment build.
### Why
In edge runtime, extract the preview props into edge functions manifest
that holding the non-deterministic inline values, then the output build
will be more deterministic that helps deployment speed
x-ref: https://github.com/vercel/vercel/pull/11390
x-ref: https://github.com/vercel/vercel/pull/11395
Closes NEXT-3012
Closes NEXT-1912
### What?
This adds a script that first gets a list of changed files compared to
our base branch (`canary`) in PRs and identifies related test files by
looking them up from a `related-tests-manifest.json`, then runs them.
### Why?
We would like to be able to run related E2E tests when certain files
change, so we can test deployment stability.
NOTE: We still want to run all E2E tests on PRs, but this script can be
used to also test deployments for hot paths. Since these tests can take
up to [3-4
hours](https://github.com/vercel/next.js/actions/workflows/test_e2e_deploy.yml),
we aim to run a subset of these tests first.
Closes NEXT-2947
Building on #58293, this expands escaping of url-unsafe characters in
paths to “required” scripts and styles in the app renderer.
This also refactors the test introduced in #58293 and expands it to
include stylesheet references as well as checking resources in the head,
which include special characters like turbopack references like
`[turbopack]`.
Test Plan: `TURBOPACK=1 pnpm test-dev
test/e2e/app-dir/resource-url-encoding`
Closes PACK-2911
### What
CSS imports in components that loaded by `next/dynamic` in client
components will cause the css are missing initial in
SSR, and loaded later on client side which will lead to FOUC. This PR
fixes the issue and get CSS preloaded in the SSR for dynamic components.
### Why
The CSS from client components that created through `next/dynamic` are
not collected in the SSR, unlike RSC rendering we already collect the
CSS resources for each entry so we included them in the server rendering
so the styles are availble at that time. But for client components, we
didn't traverse all the client components and collect the CSS resources.
In pages router we kinda collect all the dynamic imports and preload
them during SSR, but this approach is not able to be applied to app
router due to different architecture. Since we already have all the
dynamic imports info and their related chunks in
react-loadable-manifest, so we can do the similar "preloading" thing in
app router. We use the current dynamic module key (`app/page.js ->
../components/foo.js`) which created by SWC transform and match it in
the react loadable manifest that accessed from `AsyncLocalStorage`, to
get the css files created by webpack then render them as preload
styleshee links. In this way we can SSR all the related CSS resources
for dynamic client components.
The reason we pass down the react loadable manifest through
`AsyncLocalStorage` is that it's sort of exclude the manifest from RSC
payload as it's not required for hydration, but only required for SSR.
Note: this issue only occurred in dynamic rendering case for client
components.
### Other Changes Overview
- Change the react loadable manifest key from pages dir based relative
path to a source dir based relative path, to support cases with both
directory or only one of them
Closes NEXT-2578
Fixes#61212Fixes#61111Fixes#62940
Replacement for #64021 but only with production test
