$effect runs before $effect.pre

[gwwo]

Describe the bug

The parent's $effect runs before its $effect.pre, when the child component inserts a new input and shifts focus, which triggers expanded change in the parent.

<script>
  let expanded = $state(false);
  $effect(() => {
    expanded;
    console.log("run effect", expanded);
  });

  $effect.pre(() => {
    expanded;
    console.log("run pre-effect", expanded);
  });

  let inputBar;
  $effect(() => {
    if (expanded && inputBar) setTimeout(() => inputBar.focus(), 400);
  });

  import Component from "./Component.svelte";
</script>

{#if expanded}
  <input
    bind:this={inputBar}
    onblur={() => (expanded = false)}
    placeholder="input"
  />
  <Component></Component>
{:else}
  <button onclick={() => (expanded = true)}>expand</button>
{/if}

Component.svelte

<script>
  let showAnother = $state(false);
  let anotherInput;

  $effect(() => {
    if (showAnother && anotherInput) anotherInput.focus();
  });
</script>

<button
  onmousedown={
    (e) => e.preventDefault()
  } 
  onclick={() => showAnother = true}> show another input and switch focus
</button>

{#if showAnother}
  <input bind:this={anotherInput} placeholder="another input" />
{/if}

Reproduction

https://svelte.dev/playground/c60f01e426f642f7a06629fc0fbf6e22?version=5.34.8

Logs

System Info

playground

Severity

annoyance

[gwwo]

I realise that using tick().then(()=>anotherInput.focus()) in the child component fixes it. I guess it is the synchronously changing reactive value in effect that breaks things. But I am still not sure if this is just a dangerous anti-pattern.

[paoloricciuti]

I'm pretty sure this is basically #16072

[hmnd]

This is not fixed by #16930, so I think it's a more fundamental issue than #16072 and more akin to #16648 and neither this nor #16072 have been fixed by previous attempts such as #16631

[hmnd]

This is not fixed by #16930, so I think it's a more fundamental issue than #16072 and more akin to #16648 and neither this nor #16072 have been fixed by previous attempts such as #16631

@Rich-Harris related to your reply on #16930, see here :)