JS Event listeners

[AkroMD]

Good afternoon. When using aframe, you can observe a constant increase in js event listeners in the performance monitor inside the browser's dev tools. To do this, simply press any key on the keyboard. The problem repeats on an empty aframe project or any project from the examples section (https://aframe.io/aframe/examples/showcase/anime-UI ). I have not found any visible leaks, and most likely this is due to the monitor tracking feature, but I would like to understand what exactly causes this behavior.

[mrxz]

The "JS event listeners" metric does indeed increase without apparent reason. Even on a bare-bones pages like https://example.org it's possible to see it increase by consistently (double) clicking on the elements, despite the page having no JS at all.

The increments correspond with new entries appearing in the "Performance Tab" under the "Interactions" section. On pages with a canvas, like with A-Frame, any mouse/pointer/keyboard event seems to show up on the Performance tab, hence it being very noticeable.

Not sure what the root cause is, but it's clear that it's something internal to Chrome, either the way it handles/counts these or the Performance Monitor itself. In any case, it's not something A-Frame has control over.

[TriMoon]

In any case, it's not something A-Frame has control over.

If the amount of event listeners increases after each trigger fires (user input) then that's definitively a flaw in the JS code because the event listener should only be added once at startup and not on every invocation of a listener...

[mrxz]

If the amount of event listeners increases after each trigger fires (user input) then that's definitively a flaw in the JS code because the event listener should only be added once at startup and not on every invocation of a listener...

The increase happens even on pages that have no JS, so that is not something A-Frame has control over.

Of course, if there are places where components or other parts of A-Frame add event listeners without properly removing them, then that's a bug that needs to be addressed. Though generally event listeners are only added during initialization and subsequently removed during destruction, so any "leaks" would be on the application/component lifecycle, not user interaction.

[mrxz]

Looked a bit more into this. The bug is indeed inside Chrome, specifically in the web-vitals: GoogleChrome/web-vitals#594
It has been fixed in v5, but Chrome still bundles 4.2.4 (link). No idea how long it generally takes for new version to be included in Chrome itself.

There doesn't seem to be a way to disable the injection of web-vitals. Testing with an older version of Chrome before the web-vitals got integrated (<129) and the 'JS Event Listeners' metric does not increase with each input.