🦋 Changeset detected

Latest commit: 3a73842

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

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

Playground

pnpm add https://pkg.pr.new/svelte@17550

Hmm I changed the example to use $effect.pre instead of create subscriber. It not only fails against main, but also causes the tests following it to fail as well. Does that suggest the $effect.pre isn't getting cleaned up on destroy correctly?

err.... hmm. certainly sounds that way. hypothesis (with no evidence): the linked list is somehow getting broken, such that destroying the each block doesn't cause all the child effects to be destroyed

Oh that happens with either effect.pre or createsubscriber... Maybe something like #17541 is needed for both?

I think you're right, but on second thought, it's likely not a real concern after this pr. I think I was seeing it break other tests because the crash happens midway through reconciliation which messes with the linked list globally. Then because tests run in the same process, they would start off with a corrupt effect tree and break again, etc.

I was worried skipping non branch effects within each's process would mess with them getting cleaned up but that's taken care of further upstream in effect handling.