Vue 3.5.6 memory leaks

[ftrsoft]

Vue version

3.5.6

Link to minimal reproduction

Steps to reproduce

Update prodject to vue 3.5.6

What is expected?

Memory leaks. In 1 hour, memory consumption increased from 200 MB to 500 MB.
When I go back to version 3.5.5, everything becomes normal

What is actually happening?

System Info

No response

Any additional comments?

No response

[aurelsserban]

I experience the same behavior in my web app, unfortunately I am struggling to reproduce the memory leak using play.vuejs.org playground.

Unfortunately in my case the 3.5.5 version is not fixing the leaks either. I noticed that the 3.4.38 version is okay.

Maybe I am doing something wrong in my code, we have a big webapp using vue intensively and I really struggle to get to the root cause, all I can provide is the following:

1. The chain of retainers seems to follow the following pattern:
   
2. I suspect that somehow the VUE_SSR_SETTERS might be the root cause or something captured by a closure, but I am not sure.

I tried to debug it for about a week now but it seems I was not able to get very far.

I will provide another screenshot from another detached element chain of retainers (basically it seems to be the same 'link/path'):

[edison1105]

@ftrsoft @aurelsserban
Please follow the issue requirement and provide a minimal reproduction.

[yysanf]

currentInstance in watch scheduler Scopes
when watch executed in scope like this

const scope = effectScope();
const state = scope.run(() => ref({}));

const CompOne = defineComponent({
  setup() {
   state.value ={name: 'state'}
    scope.run(() => {
      watch(() => state.value.name, () => console.log('Do something'))
    })
    return { state };
  },
});

When components are destroyed， ComponentInstance in state.dep.subs.sub.scheduler Scopes, can not gc.

vue sfc

for example：Using pinia-plugin-persist can cause this problem

[edison1105]

@yysanf
You should scope.stop() first and then log state
I modified the reproduction. see Playground

[edison1105]

@ftrsoft @aurelsserban @yysanf
Please use those preview packages to test and confirm whether it resolves your issue.
If not, please provide a minimal reproduction.

[yyx990803]

I don't think the issues from @aurelsserban @yysanf are the same one as the original issue opened by @ftrsoft .

The subsHead caused memory leak is dev-only and fixed in 5c8b76e.

The original issue, however, is a regression according to @ftrsoft, but we cannot investigate it without a proper reproduction.