This PR fixes#30398
By default Next will polyfill some fetch APIs (Request, Response, Header and fetch) only if fetch is not found in the global scope in certain entry points. If we have a custom server which is adding a global fetch (and only fetch) at the very top then the rest of APIs will not be polyfilled.
This PR adds a test on the custom server where we can add a custom polyfill for fetch with an env variable. This reproduces the issue since next-server.js will be required without having a polyfill for Response which makes it fail on requiring NextResponse. Then we remove the code that checks for subrequests to happen within the **sandbox** so that we don't need to polyfill `next-server` anymore.
The we also introduce an improvement on how we handle relative requests. Since #31858 introduced a `port` and `hostname` options for the server, we can always pass absolute URLs to the Middleware so we can always use the original `nextUrl` to pass it to fetch. This brings a lot of simplification for `NextURL` since we don't have to consider relative URLs no more.
## Bug
- [x] Related issues linked using `fixes #number`
- [x] Integration tests added
- [x] Errors have helpful link attached, see `contributing.md`
This ensures type checking passes correctly for middleware types when `skipLibCheck: false` is set in `tsconfig.json`. This also moves the `middleware-types` to be an isolated test to ensure it isn't relying on any monorepo dependencies.
## Bug
- [x] Related issues linked using `fixes #number`
- [x] Integration tests added
- [ ] Errors have helpful link attached, see `contributing.md`
Fixes: https://github.com/vercel/next.js/issues/31992#issuecomment-984174872
* fix(types): add missing ua type for NextRequest
add missing ua string type on NextRequest["ua"], returns full agent as string
* Update UserAgent type
Co-authored-by: JJ Kasper <jj@jjsweb.site>
This allows TypeScript users to have type safety for middleware functions.
- Closes#30490
Co-authored-by: Lee Robinson <9113740+leerob@users.noreply.github.com>
Co-authored-by: JJ Kasper <22380829+ijjk@users.noreply.github.com>
To run middleware we are using a **sandbox** that emulates the web runtime and keeps a module cache. This cache is shared for all of the modules that we run using the sandbox while there are some module-level APIs that must be scoped depending on the module we are running.
One example of this is `fetch` where we want to always inject a special header that indicate the module that is performing the fetch and use it to avoid getting into infinite loops for middleware. For those cases the cached implementation will be the first one that instantiates the module and therefore we can actually get into infinite loops. This is the reason why #31800 is failing.
With this PR we refactor the sandbox so that the module cache is scoped per module name. This means that one execution of a middleware will preserve its cache only for that module so that each execution will still have its own `fetch` implementation, fixing this issue. Also, with this refactor the code is more clear and we also provide an option to avoid using the cache.
## Bug
- [ ] Related issues linked using `fixes #number`
- [ ] Integration tests added
- [ ] Errors have helpful link attached, see `contributing.md`
## Feature
- [ ] Implements an existing feature request or RFC. Make sure the feature request has been accepted for implementation before opening a PR.
- [ ] Related issues linked using `fixes #number`
- [ ] Integration tests added
- [ ] Documentation added
- [ ] Telemetry added. In case of a feature if it's used or not.
- [ ] Errors have helpful link attached, see `contributing.md`
## Documentation / Examples
- [ ] Make sure the linting passes by running `yarn lint`
Previously `response.cookie(name, value, options)` would mutate the passed in `options` which lead to unexpected behaviour as described in #31666.
This PR clones the `options` argument before mutating it.
## Bug
- [x] Related issues linked using `fixes #number`
- [ ] Integration tests added
- [ ] Errors have helpful link attached, see `contributing.md`
## Feature
- [ ] Implements an existing feature request or RFC. Make sure the feature request has been accepted for implementation before opening a PR.
- [x] Related issues linked using `fixes #number`
- [ ] Integration tests added
- [ ] Documentation added
- [ ] Telemetry added. In case of a feature if it's used or not.
- [ ] Errors have helpful link attached, see `contributing.md`
## Documentation / Examples
- [x] Make sure the linting passes by running `yarn lint`
Since `localhost` is actually an alias for `127.0.0.1` that points to loopback, we should take that into consideration at `NextURL` when we handle local URLs.
The implementation is based on [is-localhost-url](https://github.com/Kikobeats/is-localhost-url); I added some tests over local URLs variations present at the library to ensure other variations are working fine.
Additionally, I refactor some things over the code to avoid doing the same twice and added some legibility that is always welcome when you are working with URLs stuff.
closes https://github.com/vercel/next.js/issues/31533
This PR removes the class wrapper that we have for `TextEncoder` and `TextDecoder`. Since this is merged we will be using directly the Node version in the sandbox.
closes: https://github.com/vercel/next.js/issues/30353
According with spec, `'about:client'` is the default value is the user doesn't provide it.
It needs to add a test there, looks like there no unit tests for these classes 🤔
This fixes our `basePath` detection/replacing server-side as we were incorrectly considering `/docss` a match for a `basePath` of `/docs` which caused us to have an unexpected value in the `normalizeLocalePath` function.
- Fixes#22429
- Regression introduced in #17757
- Fixes: https://github.com/vercel/next.js/issues/31423
Co-authored-by: JJ Kasper <22380829+ijjk@users.noreply.github.com>
The `resource` argument[^1] in fetch may also be an instance of URL (or any other stringifiable value) but the sandbox variant of middlewares doesn't support that.
```js
export async function middleware(req, ev) {
await fetch(new URL('https://www.googleapis.com/oauth2/v3/certs'), {
redirect: 'manual',
method: 'GET',
})
return new Response(JSON.stringify({}), { status: 200 });
}
```
This is fixing the use of e.g. URL instance in `fetch`.
```
TypeError: initurl.startsWith is not a function
at getFetchURL (/my-next-app/node_modules/next/dist/server/web/sandbox/sandbox.js:246:17)
at fetch (/my-next-app/node_modules/next/dist/server/web/sandbox/sandbox.js:77:29)
at Object.middleware [as handler] (webpack-internal:///./pages/_middleware.js:86:15)
at async adapter (webpack-internal:///./node_modules/next/dist/server/web/adapter.js:30:22)
at async DevServer.runMiddleware (/my-next-app/node_modules/next/dist/server/next-server.js:430:26)
at async DevServer.runMiddleware (/my-next-app/node_modules/next/dist/server/dev/next-dev-server.js:394:28)
at async Object.fn (/my-next-app/node_modules/next/dist/server/next-server.js:807:34)
at async Router.execute (/my-next-app/node_modules/next/dist/server/router.js:211:32)
at async DevServer.run (/my-next-app/node_modules/next/dist/server/next-server.js:1115:29)
at async DevServer.run (/my-next-app/node_modules/next/dist/server/dev/next-dev-server.js:440:20)
```
[^1]: https://developer.mozilla.org/en-US/docs/Web/API/fetch#parameters
When libraries are required outside of the middleware function context and they do checks such as `a instanceof Uint8Array` since the constructors are different between the two contexts they'll always yield false.
This is a problem for libraries validating user input as well as the WebCryptoAPI polyfill used outside of Edge Functions.
- Fixes#30477
- Fixes#30911
This is only a problem for the sandbox runtime, not when ran inside an Edge Function.
Co-authored-by: Tobias Koppers <sokra@users.noreply.github.com>
With this PR we are updating the way we check the usage of `eval` and other dynamic code evaluation (like `new Function`) for middleware. Now instead of simply showing a warning it will behave differently depending on if we are building or in development.
- Development: we replace the dynamic code with a wrapper so that we print a warning only when the code is used. We don't fail in this scenario as it is possible that once the application is built the code that uses `eval` is left out.
- Build: we detect with tree shaking if the code that will be bundled into the middleware includes any dynamic code and in such scenario we make the build fail as don't want to allow it for the production environment.
Closes#30674
## Bug
- [ ] Related issues linked using `fixes #number`
- [ ] Integration tests added
- [ ] Errors have helpful link attached, see `contributing.md`
## Feature
- [ ] Implements an existing feature request or RFC. Make sure the feature request has been accepted for implementation before opening a PR.
- [ ] Related issues linked using `fixes #number`
- [ ] Integration tests added
- [ ] Documentation added
- [ ] Telemetry added. In case of a feature if it's used or not.
- [ ] Errors have helpful link attached, see `contributing.md`
## Documentation / Examples
- [ ] Make sure the linting passes by running `yarn lint`
- Code splitting should be disabled for the server-web build. Done via `ServerlessPlugin`.
- ~Target can't be `web`, `webworker` is better.~ Using `web` and `es6` for now, still not ideal.
- https://github.com/acornjs/acorn/issues/970
## Bug
- [ ] Related issues linked using `fixes #number`
- [x] Integration tests added
- [ ] Errors have helpful link attached, see `contributing.md`
## Feature
- [ ] Implements an existing feature request or RFC. Make sure the feature request has been accepted for implementation before opening a PR.
- [ ] Related issues linked using `fixes #number`
- [ ] Integration tests added
- [ ] Documentation added
- [ ] Telemetry added. In case of a feature if it's used or not.
- [ ] Errors have helpful link attached, see `contributing.md`
## Documentation / Examples
- [ ] Make sure the linting passes by running `yarn lint`
## Bug
- [x] Related issues linked using `fixes #number`
- [x] Integration tests added
- [ ] Errors have helpful link attached, see `contributing.md`
Fixes#30430
There's some more discussion in the issue, but in summary:
- web `Headers` implementation combines all header values with `', '`
- For `Set-Cookie` headers, you're supposed to set them as separate values, not combine them
- web `Headers` forbids the use of `Cookie`, `Set-Cookie` and some more headers, so they don't have custom implementation for those, and still joins them with `,`
- We currently just split them using `split(',')`, but this breaks when the header contains a date (expires, max-age) that also includes a `,`
I used this method to split the Set-Cookie header properly: https://www.npmjs.com/package/set-cookie-parser#splitcookiestringcombinedsetcookieheader as suggested [here](https://github.com/whatwg/fetch/issues/973#issuecomment-559678813)
I didn't add it as a dependency, since we only needed that one method and I wasn't sure what the process is for adding dependencies, so I just added the method in the middleware utils
This PR adds support for [Middleware as per RFC ](https://github.com/vercel/next.js/discussions/29750).
## Feature
- [ ] Implements an existing feature request or RFC. Make sure the feature request has been accepted for implementation before opening a PR.
- [ ] Related issues linked using `fixes #number`
- [ ] Integration tests added
- [ ] Documentation added
- [ ] Telemetry added. In case of a feature if it's used or not.
- [ ] Errors have helpful link attached, see `contributing.md`
## Documentation / Examples
- [ ] Make sure the linting passes