[renovate]

This PR contains the following updates:

Package	Change	Age	Confidence
@astrojs/cloudflare (source)	14.3.4 → 14.4.0
@astrojs/react (source)	7.0.1 → 7.1.0
astro (source)	^7.3.8 → ^7.4.0
undici (source)	8.12.1 → 8.12.2

⚠️ Renovate does not enforce Minimum Release Age for bump, lockfileUpdate, or rollback updates, so these are raised without a Minimum Release Age check. You will need to manually validate the Minimum Release Age for these package(s).

---

Release Notes

withastro/astro (@​astrojs/cloudflare)

v14.4.0

Compare Source

Minor Changes

• #​18177 7d814e5 Thanks @​Princesseuh! - Reworks image processing on prerendered pages to use Astro's built-in image pipeline. Requires astro@^7.4.0

  Images on prerendered pages now behave the same as on other adapters. Images optimized with the Cloudflare Images binding are now properly cached between builds, and image options such as assetQueryParams and build.assetsPrefix are applied the same way as everywhere else.

Patch Changes

• Updated dependencies [b7a1bd9]:
  • @​astrojs/internal-helpers@0.13.0
  • @​astrojs/underscore-redirects@1.0.4

withastro/astro (@​astrojs/react)

v7.1.0

Compare Source

Minor Changes

• #​18277 2715958 Thanks @​florian-lefebvre! - Increases the minimum Node.js version to 22.22.0

  The minimum supported Node.js version is now 22.22.0, the latest Node.js 22 release supported by StackBlitz. This change is in line with Astro's Node.js support policy.

Patch Changes

• Updated dependencies [b7a1bd9]:
  • @​astrojs/internal-helpers@0.13.0

withastro/astro (astro)

v7.4.0

Compare Source

Minor Changes

• #​17894 b7a1bd9 Thanks @​Princesseuh! - Adds a new --timings flag to astro build to help you find what slows down your builds

  When a build takes longer than expected, it can be hard to tell whether the cause is an integration, a Markdown processor plugin, image processing, or even syntax highlighting. Add the --timings flag to your build command to print a summary once the build finishes:

  astro build --timings

  This is helpful to decide what to optimize first, or to compare build times before and after a change to your project.

• #​17830 a937b0d Thanks @​florian-lefebvre! - Adds a new optional getFontProperties() callback to font providers

  Font providers can now expose the weights, styles, subsets and formats a font family actually has. Astro uses it when resolving your font configuration to warn about values a provider cannot serve, for example a weight a font family does not offer.

  All built-in providers implement it. If you maintain a custom font provider, implementing this callback is optional for now but it will be required in Astro 8:

  import type { FontProvider } from 'astro';
  
  const provider: FontProvider = {
  	name: 'my-provider',
  	resolveFont: ({ familyName }) => {
  		// ...
  	},
  	getFontProperties: ({ familyName }) => ({
  		weights: ['400', '700'],
  		styles: ['normal', 'italic'],
  		subsets: ['latin'],
  		formats: ['woff2'],
  	}),
  };

• #​17830 a937b0d Thanks @​florian-lefebvre! - Adds a fetch() function to the font provider init() context

  This function requests a provider API, retrying transient failures and more. It lets providers that wrap a 3rd-party unifont provider pass the context to it directly:

  import type { FontProvider } from 'astro';
  import type { InitializedProvider } from 'unifont';
  import { acmeProvider } from '@acme/unifont-provider';
  
  export function acmeFontProvider(): FontProvider {
  	const provider = acmeProvider();
  	let initializedProvider: InitializedProvider | undefined;
  	return {
  		name: provider._name,
  		async init(context) {
  			initializedProvider = await provider(context);
  		},
  		// ...
  	};
  }

• #​18015 0b30239 Thanks @​florian-lefebvre! - Adds experimental.redirectPage, letting you provide your own src/pages/3xx.astro in place of the redirect page Astro generates in static builds.

  A static site has no server to send an HTTP Location header, so Astro generates a small HTML page with a <meta http-equiv="refresh"> tag for each redirect it builds:

  • Your redirects entries
  • Astro.redirect() calls in prerendered pages
  • i18n redirects

  That page's markup, wording, and refresh delay were fixed, so you could not brand or translate it.

  To enable this feature, add the experimental flag to your Astro config:

  // astro.config.mjs
  import { defineConfig } from 'astro/config';
  
  export default defineConfig({
    experimental: {
      redirectPage: true,
    },
  });

  Then create src/pages/3xx.astro. Astro renders it once per redirect, passing that redirect's details as props:

  ---
  // src/pages/3xx.astro
  import type { RedirectPageProps as Props } from 'astro';
  
  const { from, to, status, delay } = Astro.props;
  ---
  
  <html lang="en">
    <head>
      <meta http-equiv="refresh" content={`${delay};url=${to}`} />
      <meta name="robots" content="noindex" />
      <title>Redirecting…</title>
    </head>
    <body>
      <p>This page has moved. <a href={to}>Continue to {to}</a>.</p>
    </body>
  </html>

  Your page must render a <meta http-equiv="refresh"> tag pointing at to, because in a static build that tag is the redirect. Astro logs a warning during the build if it is missing.

  See the experimental redirect page documentation for more information.

• #​18265 910ee69 Thanks @​astro-factory! - Adds a notFound() handler to astro/fetch and astro/hono that renders the custom 404 page when no route matches, including a prerendered 404.astro

  Add it before pages() in a custom worker or server entrypoint that does not run inside App.render, such as a Hono app used directly as a Cloudflare worker. Without it, pages() and middleware() return an empty 404 with an internal X-Astro-Error header for unmatched routes. astro() already includes this handling.

  import { Hono } from 'hono';
  import { notFound, pages } from 'astro/hono';
  import { cf } from '@astrojs/cloudflare/hono';
  
  const app = new Hono();
  app.use(cf());
  app.use(notFound());
  app.use(pages());
  export default app;

• #​18177 7d814e5 Thanks @​Princesseuh! - Adds support for separate build and runtime image services

  image.service now also accepts a build and a runtime service. The build service optimizes images on prerendered pages when you build your site, and is also used during development. The runtime service handles images on on-demand rendered pages and the image endpoint of your deployed site.

  The following example uses Sharp to optimize images on prerendered pages, and an image CDN for on-demand rendered pages:

  // astro.config.mjs
  import { defineConfig } from 'astro/config';
  
  export default defineConfig({
    image: {
      service: {
        build: { entrypoint: 'astro/assets/services/sharp' },
        runtime: { entrypoint: 'my-image-cdn-service' },
      },
    },
  });

Changes for integration authors

config.image.service keeps the shape the user configured it with. When a project sets build and runtime services, config.image.service.entrypoint and config.image.service.config are undefined.

If your integration reads image.service, check for the runtime property to get the service you need:

'astro:config:done': ({ config }) => {
  const { service } = config.image;
  const runtimeService = 'runtime' in service ? service.runtime : service;

  if (runtimeService.entrypoint === 'astro/assets/services/sharp') {
    // ...
  }
},

Setting image.service with updateConfig() replaces the existing service instead of merging with it when the two use different shapes. For example, setting a single service in a project that uses build and runtime services replaces both.

• #​18277 2715958 Thanks @​florian-lefebvre! - Increases the minimum Node.js version to 22.22.0

  The minimum supported Node.js version is now 22.22.0, the latest Node.js 22 release supported by StackBlitz. This change is in line with Astro's Node.js support policy.

• #​18313 71e2b24 Thanks @​matthewp! - Allows integrations to control where Astro writes client, server, and prerender build output.

  Custom prerenderers can access the final output locations through the new outputDirectories context:

  setPrerenderer((defaultPrerenderer, { outputDirectories }) => {
    const { client, server, prerender } = outputDirectories
  
    return createPrerenderer({
      defaultPrerenderer,
      client,
      server,
      prerender,
    })
  })

• #​18177 7d814e5 Thanks @​Princesseuh! - Lets custom prerenderers report the images used on prerendered pages

  Astro now generates images on prerendered pages from the image information that a prerenderer's getStaticPaths() and render() return. If your prerenderer wraps Astro's default one, pass this information through:

  setPrerenderer((defaultPrerenderer) => ({
    name: 'my-prerenderer',
    async setup() {
      await defaultPrerenderer.setup?.();
    },
    async getStaticPaths() {
      const result = await defaultPrerenderer.getStaticPaths();
      const { paths, metadata } = Array.isArray(result) ? { paths: result } : result;
      return { paths: paths.filter(({ pathname }) => !pathname.startsWith('/drafts')), metadata };
    },
    render(request, options) {
      return defaultPrerenderer.render(request, options);
    },
    async teardown() {
      await defaultPrerenderer.teardown?.();
    },
  }));

  Returning a plain array from getStaticPaths() or a plain Response from render() still works, but is deprecated: images used on those pages are not generated.

• #​17932 b905ff1 Thanks @​Princesseuh! - Adds support for experimental parallel prerendering

  This experimental feature makes astro build render your static pages in parallel, using multiple CPU cores at once instead of one. On sites with many pages, this can significantly reduce build times, at the cost of higher memory usage.

  To enable this feature, add the experimental flag in your Astro config:

  // astro.config.mjs
  import { defineConfig } from 'astro/config';
  
  export default defineConfig({
    experimental: {
      parallelPrerender: true,
    },
  });

  By default, Astro uses all available CPU cores but one. Each additional worker increases the memory used by your build. If your build runs out of memory, or does not get faster, set workers to a lower number:

  // astro.config.mjs
  import { defineConfig } from 'astro/config';
  
  export default defineConfig({
    experimental: {
      parallelPrerender: {
        workers: 4,
      },
    },
  });

  Each worker renders as many pages at once as your build.concurrency setting allows.

  This feature has no effect when your adapter provides its own way of prerendering pages. Among official adapters, this only applies to @astrojs/cloudflare. Astro logs a warning during the build when the flag is ignored.

  For more information on enabling and using this feature in your project, see the experimental parallel prerendering docs.

Patch Changes

• #​18313 71e2b24 Thanks @​matthewp! - Fixes deployment integrations being unable to use Vite manifests when preparing build output

• #​18270 0c3bf11 Thanks @​ArmandPhilippot! - Fixes the JSDoc for CSP utilities to point to the render context reference in docs.

• #​18368 1f04637 Thanks @​43081j! - Replaces internal use of find-proc with tinyps.

• #​18355 59b9cc0 Thanks @​matthewp! - Fixes prerender errors being silently swallowed when using a custom src/fetch.ts with an adapter that prerenders in its own runtime, such as @astrojs/cloudflare. Previously, a page that threw during prerendering was written as an empty HTML file and the build succeeded. The build now fails with the underlying error.

• #​18177 7d814e5 Thanks @​Princesseuh! - Fixes original images being kept despite only width or height being read.

• #​18353 af54bfa Thanks @​astro-factory! - Fixes glob() loader swallowing rendering errors during content sync, which allowed astro build to exit successfully (code 0) even when a remark/rehype plugin threw an error

• #​18271 23e8e28 Thanks @​ArmandPhilippot! - Fixes the sync command description in astro help, which only mentioned content collections.

• Updated dependencies [b7a1bd9, 2715958]:

  • @​astrojs/internal-helpers@0.13.0
  • @​astrojs/markdown-satteri@0.4.4
  • @​astrojs/telemetry@3.4.0

nodejs/undici (undici)

v8.12.2

Compare Source

Security fixes

High severity

Denial of service via TLS 1.2 certificate-changing renegotiation

CVE-2026-87018 · GHSA-qrxm-fqhp-jwvm

A malicious or compromised HTTPS server could crash the Node.js process by initiating TLS 1.2 renegotiation and presenting a different, CA-trusted certificate whose hostname did not match the requested host. The resulting socket error reached an assertion in undici's HTTP/1.1 and HTTP/2 connection handling. This release handles post-connect certificate errors without terminating the process.

TLS certificate validation bypass through dropped Agent options

CVE-2026-109383 · GHSA-2m5v-pch6-74hx

Agent used a JSON-based clone for constructor options, which silently removed function-valued TLS options such as checkServerIdentity. Applications relying on custom certificate or public-key verification could therefore accept a certificate that the callback was intended to reject. This release preserves TLS options when creating per-origin dispatchers.

Cache poisoning and cross-user disclosure through cross-origin redirects

CVE-2026-109829 · GHSA-4w7w-5xrc-9h26

When an originless Agent combined redirect and shared-cache interceptors as compose(redirect(), cache()), a response from a cross-origin redirect could be stored under the original URL. This could allow persistent cache poisoning or disclosure of user-specific responses to later callers. This release verifies the response origin before caching and preserves redirect-hop cache-control policy.

Denial of service through excessive Content-Encoding values

CVE-2026-109869 · GHSA-xjj6-f33q-929c

A malicious HTTP server could crash the Node.js process by returning more than five content codings and closing the connection before sending the body. Although fetch() rejected the response, an error from an orphaned internal stream could remain unhandled. This release validates the content-encoding chain before creating the body stream and aborts oversized chains safely.

Medium severity

Denial of service through the decompress interceptor

CVE-2026-87029 · GHSA-55wc-3pg8-jfw6

An exception thrown by a downstream response handler from an asynchronous decompression callback could escape as an uncaught exception. A server could trigger this when decompress was composed with a handler such as dump with a size limit. This release converts downstream handler exceptions into per-request errors.

Subresource Integrity bypass with non-lowercase algorithm tokens

CVE-2026-109841 · GHSA-f72v-6w45-q472

fetch() treated valid mixed-case or uppercase SRI algorithm tokens, such as SHA256, as unknown and could accept a response whose body did not match the configured digest. This release normalizes algorithm tokens before validation and digest selection.

Denial of service through unbounded EventSource parser state

CVE-2026-109924 · GHSA-4qv8-2p9j-2685

A malicious Server-Sent Events endpoint could exhaust process memory by sending many tiny data: fields without completing an event, or by splitting an unfinished line across many chunks. This release bounds retained EventSource parser state, including field counts, buffered chunks, unfinished-line length, and retained field data.

Denial of service through unbounded automatic WebSocket PONG buffering

CVE-2026-109932 · GHSA-9fm3-97px-rjf6

A malicious WebSocket server could exhaust process memory by flooding the client with PING frames while not reading the automatic PONG responses. undici previously ignored socket backpressure and allowed the outbound queue to grow without bound. This release pauses frame parsing when PONG writes encounter backpressure and resumes after the socket drains.

Full Changelog: nodejs/undici@v8.12.1...v8.12.2

---

Configuration

📅 Schedule: (UTC)

• Branch creation
  • Only on Sunday and Saturday (* * * * 0,6)
• Automerge
  • At any time (no schedule defined)

🚦 Automerge: Enabled.

♻ Rebasing: Whenever PR is behind base branch, or you tick the rebase/retry checkbox.

👻 Immortal: This PR will be recreated if closed unmerged. Get config help if that's undesired.

---

• If you want to rebase/retry this PR, check this box

---

This PR was generated by Mend Renovate. View the repository job log.

[sonarqubecloud]

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud