fix(slots): check nullability for multi-branch dynamic slots

[shigma]

Check nullability for multi-branch dynamic slots, fix ssr regression bug introduced in 3.2.38.

Fix #6639
Fix #6651

[posva]

can you add a unit test?

[shigma]

Note: I'm not sure the fix touches the the core of problem (although it indeed fixes the regression introduced in 00036bb).

type Slot = (...args: any[]) => VNode[]

From the above typings, a Slot should not return undefined. However it does in the simple reproduction #6639. This means there may be other bugs in the repo but I could not find them (or maybe we need to change the typings).

Despite this, I still think it is necessary to check nullability before we could locate the original source of the problem and even we could, it is still uncertain that there is no more slots returning undefined. After all, this regression will affect all SSR builds which include dynamic slots.

[yyx990803]

Thanks for the PR, sorry I didn't notice it when pushing 8963c55 - I'm merging this for the unit test.

Regarding the return type - the return value is undefined when the slots are SSR-compiled. In SSR mode the slots have two branches based on whether the child component consuming the slots is also SSR-compiled or vdom-based. The branches are detected based on whether there are SSR-specific helpers being passed in when the slot is called, so this is strictly an internal case and only affect internal runtime helpers that work with slots and are used in both DOM and SSR envs. I've checked the type usage and there are no other helper that falls into this category.