Use suspense boundary vnode as parent for subtree re-render

[f0x52]

Fixes #406

[changeset-bot]

🦋 Changeset detected

Latest commit: 2022f10

The changes in this PR will be included in the next version bump.

This PR includes changesets to release 1 package

Name	Type
preact-render-to-string	Patch

Not sure what this means? Click here to learn what changesets are.

Click here if you're a maintainer who wants to add another changeset to this PR

[JoviDeCroock]

This does look good, we'll need a changeset and a few tests to prevent regressions. You might need to update the types of the renderer as well

[f0x52]

Would this be a patch changeset?

[f0x52]

Added changeset, updated types, and added a single test that traverses VNode parents inside the suspense boundary and checks if there is a circular reference.
I couldn't find a good way to add a test that actually tries to use the useId hook, because in case of a regression this will cause the tests to run forever; since the Node.js event loop is continually busy inside the while loop, Mocha's timeout functionality can't stop the test either (mochajs/mocha#1609)

[JoviDeCroock]

You should do the useId test and leverage the abortController to abort after i.e 1 second and fail

[f0x52]

I don't think that's possible either? The synchronous while loop inside the useId hook completely blocks the Node.js event loop, so as far as I know there is no way to abort the abortController while it's in the loop; timeouts/intervals won't even be called. Tripwire as mentioned in the mocha issue uses a native extension to be able to interrupt unbreakable loops, but including a native extension + build step doesn't seem ideal either.

Perhaps it would be good to add a check in while loops that could otherwise run forever on erroneous node trees; so far I found this in useId, and the _catchError handler, though that's probably a performance tradeoff

[JoviDeCroock]

I mean, it works now so we should not care about a test loop.... we should write a test, i can write it if you prefer.

[f0x52]

This test also expects the current behavior where useId's are scoped to the rendering root/boundary, resulting in id P0-0 for the first element outside the suspense boundary, and P0-0 for the first element inside as well.