router.load() resolves before an ancestor boundary's code-split errorComponent is loaded (regression in 1.170.19)
维护者通常 1 天内回复
评估
- 难度
- 4/5
- 预计耗时
- 1-2 天
- 新手友好度
- 55/100
- Issue 类型
- 缺陷
- 描述清晰度
- 基本清楚
- 活跃度
- 活跃
- 技术栈
- react, typescript, vite
- 领域
- frontend
调研方向
Start by tracing router.load() and route chunk loading, especially how an errored child selects an ancestor’s errorComponent; the report points to a regression introduced in 1.170.19 and possibly related to #7926. Use the linked reproducer and its failing test, then compare behavior with 1.170.18. Done means the boundary component is loaded when router.load() resolves and the regression test passes.
由索引模型根据 Issue 内容生成。
描述
Which project does this relate to?
Router
Describe the bug
Since 1.170.19, router.load() can resolve before the chunk for the errorComponent that will actually render has loaded. As a result, the first render after load() suspends and is empty.
This happens when a child's loader throws, the child has no errorComponent, and an ancestor's errorComponent is code-split (which autoCodeSplitting does by default). The errored match only loads its own errorComponent, and the ancestor is success so it only loads component. So nothing preloads the boundary that catches the error. Before 1.170.19, every matched route's chunk loaded all of its component types, so this worked.
const parent = createRoute({
getParentRoute: () => rootRoute,
path: '/parent',
errorComponent: lazyRouteComponent(() => import('./ParentError'), 'ParentError'),
component: () => <Outlet />,
})
const child = createRoute({
getParentRoute: () => parent,
path: '/child',
loader: () => {
throw new Error('child loader failed')
},
component: () => <p>Child</p>,
})
await router.load()
flushSync(() => createRoot(el).render(<RouterProvider router={router} />))
// 1.170.18: el.textContent === 'Parent errorComponent'
// 1.170.46: el.textContent === '', the error component shows up once the chunk arrives
In the browser this is an empty frame. Under Testing Library (IS_REACT_ACT_ENVIRONMENT), the suspended render inside sync act() is never retried, so a test that awaits router.load() and then renders sees a blank page forever. That's how I found it.
Complete minimal reproducer
https://github.com/loganrosen/tanstack-router-load-lazy-error-component
Steps to Reproduce the Bug
pnpm install && pnpm dev, then open the page. The log at the top shows what the first commit afterrouter.load()contained.- On 1.170.46 it logs
first commit: ""and then+312.0ms: "Parent errorComponent". - Set
@tanstack/react-routerto1.170.18,pnpm install, reload. It logsfirst commit: "Parent errorComponent". pnpm testfails on 1.170.46 and passes on 1.170.18.
Expected behavior
When router.load() resolves, the errorComponent/notFoundComponent of the boundary that will render the error should already be loaded, like it was before 1.170.19. The API docs say load() "resolves when they are all loaded and ready to be rendered".
Screenshots or Videos
No response
Platform
- Router / Start Version: @tanstack/react-router 1.170.46 (router-core 1.171.39); works on 1.170.18 (router-core 1.171.15)
- OS: macOS
- Browser: WebKit (macOS system webview)
- Bundler: vite
- Bundler Version: 8.3.5
Additional context
Found while upgrading the router in matrix-authentication-service (element-hq/matrix-authentication-service#6069). We're working around it in our test helper by calling router.loadRouteChunk(route, 'errorComponent') for every match after load(). Possibly related to the loader rewrite in #7926.
The issue was written with help from an AI coding assistant. I ran and checked the reproducer myself.
- 主要语言
- TypeScript
- 星标
- 15.3k
- 派生
- 1.9k
- 平均合并
- 1 天 5 小时
- 30 天内合并 PR
- 46
环境准备
在浏览器里用你自己的 GitHub 账号启动这个项目的开发容器。
- 没有 Dockerfile 或 Docker Compose 文件
- 有 Pull Request 模板
- 阅读贡献指南
从这里开始
- 先读完整个 Issue,再读项目的贡献指南。
- 在 Issue 下留言说明你要接手 —— 这能避免两个人做同样的事。
- Fork 仓库,在一个分支上完成修改。
- 提交 Pull Request,并在描述里引用这个 Issue 编号。
TanStack/router 的其他 Issue
-
难度 2/5 1-3 小时 新手友好度 85/100
维护者通常 1 天内回复
-
Masked navigation drops the first character of the hash each time (__tempLocation hash missing '#')可能已有人在做 @hassanabrar202 于 8 天前认领。 未关闭
难度 2/5 1-3 小时 新手友好度 72/100
维护者通常 1 天内回复
-
router-generator: lazy-load Prettier (costs ~300–450ms per build even when formatting is off)可能已有人在做 @Sheraff 于 27 天前认领。 未关闭package: router-generator performance
难度 2/5 1-3 小时 新手友好度 85/100
维护者通常 1 天内回复
-
router-generator: sibling `physical()` subtrees in a virtual route config generate duplicate identifiers可能已有人在做 @Sheraff 于 27 天前认领。 未关闭
难度 2/5 1-3 小时 新手友好度 76/100
维护者通常 1 天内回复
-
docs(router): clarify static title fallback in SPA mode可能已有人在做 @Rohan-Dedhia 于 17 天前认领。 未关闭
难度 1/5 1-3 小时 新手友好度 62/100
维护者通常 1 天内回复
相似的 Issue
-
[BUG]: Shared links drop the configured basePath on GitHub Pages可能已有人在做 @anshulchikhale30-p 今天认领。 未关闭bug
难度 1/5 1 小时以内 新手友好度 88/100
StabilityNexus/IdentityTokens-EVM-Frontend#182 · 1 条评论 ·
维护者通常 1 天内回复
-
Add: GTrekker未关闭check:passed streams:add
难度 1/5 1 小时以内 新手友好度 72/100
维护者通常 1 天内回复
-
难度 2/5 1-3 小时 新手友好度 70/100
Koenkk/zigbee-herdsman-converters#13900 ·
维护者通常 1 天内回复
-
area: ui Bug Payload OSS
难度 1/5 1 小时以内 新手友好度 88/100
payloadcms/payload#19839 ·
维护者通常 1 天内回复
-
难度 1/5 1 小时以内 新手友好度 88/100
jeanibarz/knowledge-base-mcp-server#1291 ·
维护者通常 1 天内回复