hono 4.9.8 typing regression

[c4rloscg]

What version of Hono are you using?

4.9.8

What runtime/platform is your app running on? (with version if possible)

Bun 1.2.20

What steps can reproduce the bug?

After upgrading to hono 4.9.8, TypeScript inference for the client created with hc<ApiRoutes> no longer preserves per-route response generics. In my codebase this causes generated client calls like rootClient.routeA.search.$post(...) used with tanstack/react-query@5.90.2 useInfiniteQuery to produce InfiniteData<unknown, unknown>.

This regression appears related to the changes introduced in #4405. Pinning to hono 4.9.7 restores the previous typing behavior.

• react ^19.1.0

What is the expected behavior?

hc<ApiRoutes> should create a typed client where each endpoint method exposes the correct response generic corresponding to the server route, permitting downstream code to infer the actual response shape.

What do you see instead?

No response

Additional information

No response

[yusukebe]

@c4rloscg Thank you for the issue.

Hey @BarryThePenguin, can you take a look at this?

[BarryThePenguin]

Hi @c4rloscg, can you provide a bit more detail please?

What do the routes respond with? What do you expect instead of InfiniteData<unknown, unknown>?

interface InfiniteData<TData, TPageParam = unknown> {
  pages: Array<TData>
  pageParams: Array<TPageParam>
}

From what I can see, even without the type parameters, we're still only dealing with values that can be parsed as JSON, eg;

const example: InfiniteData<{ title: string }> = {
  pages: [{ title: 'foo' }, { title: 'bar' }],
  pageParams: []
}

[aantthony]

Perhaps related to #4427?

[c4rloscg]

I will close this issue since I have not been able to reproduce this outside my project. My suspicion is that type inference is now more accurate, raising an error in my code. Using tanstack/react-query@5.90.2 useInfiniteQuery, I was returning a value in getNextPageParam typed as prop?: unknown; | prop: MyType. Removing the unknown portion with a more appropriate type removes the error.

Thank you for taking a look into this!