Release January 20, 2026 - #2582
Open
github-actions[bot] wants to merge 251 commits into
Open
Conversation
fix: use immediate input value for Enter navigation
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
fix: get redirectTo from the url params instead of the headers
github-actions
Bot
force-pushed
the
changeset-release/from-canary
branch
from
January 23, 2026 15:25
0fc4cb5 to
6629180
Compare
…ieve the flags in the middleware as well.
… for plugins. Disabledby default.
Feature/private catalog
Enabling Preview Mode in production requires typing the `secret` (it is only pre-filled in dev), which dirties the field and persists it to sessionStorage. `PreviewModeEnabled` then restored that secret via `setValue(shouldDirty)`, which the all-fields `FormAutoSubmit` read as a user change and submitted the `update` action — a `window.location` hop to `/api/preview` that 307-redirects back and re-triggers the restore, looping forever. Scope `FormAutoSubmit` to `previewData` (the only edited field) and exclude the `secret` from `FormPersist` so the token is no longer stored in the browser. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Overrides the terminating link at the tail of the Apollo link chain (by default the HttpLink to the Mesh backend). Because it runs after every context-setting link (customer auth token, store, cache-id, headers), it lets you route specific operations to a different transport while still inheriting all request headers. Motivating case: file uploads. File/Blob variables must be sent as a multipart request (apollo-upload-client's UploadHttpLink), which the default HttpLink cannot serialize. Prepending such a link via `links` terminated it before the auth/header links ran, dropping the customer token from multipart requests. Supplying the upload-aware split as terminatingLink keeps it at the tail so uploads inherit the token like any other operation. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…xt-bodyparser fix(graphql-mesh): disable Next bodyParser so multipart uploads survive the mesh
…port
findOriginalSource only matched an export specifier's `exported` field, which
SWC leaves null for a non-aliased re-export like `export { Image }` — the name
lives in `orig`, and `exported` is only set by an aliased `export { X as Y }`.
Components re-exported without an alias (e.g. `Image` from @graphcommerce/image)
were therefore unreachable by plugins and failed with "Plugin target not found".
Fall back to `orig` when `exported` is absent.
dist/index.js is committed and consumed directly (no repo build step), so the
compiled output is updated to match the src change; prepack rebuilds it cleanly
on the next release.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
…/executor-http FormData mismatch customFetch used globalThis.fetch (undici) inside Next while @graphql-tools/executor-http builds multipart bodies with FormData from @whatwg-node/fetch. In next dev (turbopack) @whatwg-node/fetch evaluates during next.config.ts loading, before any __NEXT global exists, so its Next.js detection fails and it exports ponyfills — undici then stringifies the ponyfill FormData body to the literal '[object Object]' (Magento: "Unable to parse the request."). Always use @whatwg-node/fetch's own fetch so fetch and FormData come from the same implementation family. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
…ata-fetch-mismatch fix(graphql-mesh): multipart uploads broken in next dev — customFetch/executor-http FormData mismatch
…cept-non-aliased-reexports fix(next-config): intercept components exported via non-aliased re-export
…-terminating-link feat(graphql): add terminatingLink prop to GraphQLProvider
…erceptors export * from './X.original' does not re-export default (ES semantics), so an intercepted module silently lost its default export. GraphQL Mesh resolves customFetch via `exported.default || exported`, so a plugin on @graphcommerce/graphql-mesh/customFetch handed Mesh the module namespace instead of the fetch function, breaking schema introspection during gc-mesh build. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
…rd-default-export fix(next-config): forward the original module's default export in interceptors
…nite-reload fix(ecommerce-ui): stop Preview Mode infinite reload in production
…a GC_MAGENTO_ENDPOINT_SERVER When the runtime environment sets GC_MAGENTO_ENDPOINT_SERVER, customFetch rewrites every request matching the origin of GC_MAGENTO_ENDPOINT to the internal origin (GraphQL and REST) and adds X-Forwarded-Proto: https. This keeps frontend<->Magento traffic inside the cluster instead of hairpinning over the public load balancer. Runtime-only by design: build-time schema introspection and SSG keep using the public endpoint. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
…e errors to the customer inside checkout / Added card icon for the method
A `<video>` paints nothing until it has buffered enough for its first
frame, and nothing at all when autoplay is blocked (iOS Low Power Mode).
An autoplaying video banner therefore starts out black, and can stay
black. `Asset` now takes a `poster`, rendered as `<video poster>`.
Storyblok's `type: asset` has no room for a poster, so `assetWithPoster()`
reads the convention of an asset value carrying an extra `poster` key:
{ "fieldtype": "asset", "filename": "…", "poster": { "filename": "…" } }
Keeping the poster beside the asset rather than nesting both under a
wrapper is what makes a custom field type storing that shape a drop-in
for a plain asset field: existing content stays valid and
`value.filename` keeps working for consumers that ignore the poster. The
narrowing is unavoidable regardless — Storyblok has no JSONSchema for
custom field types, so its type generator emits `unknown` for them.
`@graphcommerce/image` gains `imageUrl()`, which builds an optimized URL
outside of a React tree for the spots that need a bare URL rather than an
`<Image>` — `<video poster>`, `background-image`, `og:image`. It routes
through the configured loader, so the bytes are served and cached by the
deployment instead of every visitor hitting the origin host; that matters
when the origin meters bandwidth. Width snaps up to the nearest
configured size, since the optimizer 400s on anything outside
`imageSizes`/`deviceSizes`.
Improve Braintree checkout flow and make 3D Secure optional
…ndpoint feat(graphql-mesh): route Magento traffic over an internal network via GC_MAGENTO_ENDPOINT_SERVER
…Poster Each asset field in the example now goes through `assetWithPoster()` instead of reading `blok.asset` directly, and forwards the resulting `poster` to `<Asset>`. This makes the choice of field type a content decision rather than a code one: a plain `type: asset` and a custom field storing the poster convention both narrow through the same call, so a project can switch any field between the two — in either direction — without touching a component. Fields that never hold a video simply resolve `poster` to undefined.
`HeroBanner` rendered a `<video>` itself, from a raw URL. That meant it
could only ever hold a video, and left nowhere to hang a poster — the one
component where a poster matters most, being a full-bleed autoplaying
video above the fold.
It now takes `asset?: React.ReactNode` and renders it, exactly as its
sibling SpecialBanner already does. Positioning stays with the banner:
whatever is passed gets stretched to fill via `& img, & video`.
- <HeroBanner videoSrc={asset.filename} … />
+ <HeroBanner asset={<Asset asset={asset} poster={poster} />} … />
Taking a node rather than data also keeps next-ui free of any CMS: the
Storyblok and Hygraph examples each pass their own `<Asset>`, both of
which already render images and video. Importing a CMS `Asset` into
next-ui would have inverted the dependency — storyblok-ui and hygraph-ui
depend on next-ui, not the other way round.
Non-breaking: `videoSrc` stays as a deprecated fallback that renders a
bare autoplaying video (keeping the HeroBanner-video class), and the
`animated`/`video` class parts stay exported. The only behavioural change
on the videoSrc path is the loss of the scroll parallax, along with the
framer-motion, useScrollY and clientSize machinery it required.
…et-poster Feature/storyblok asset poster
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
5 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Automated release PR from canary branch