vg.no was addressed here: 0dc0d44

Better to address these sites separately, like at #22295

Yesterday I had tried that sort of filter shortly after my other proposed ones in the other thread:

voelkner.de##+js(trusted-click-element, 'button[data-cookie_consent="0"]')

However I found out that a delay was needed otherwise this doesn't work. Stepping through the code, I see that without delay, the mutation observer somehow is not triggered after the DOM is finally filled with the dialog element.

So I have changed the code to no longer rely on mutation observer but rather on requestIdleCallback to wait for the target element to become part of the DOM. This allows the current case to be fixed without requiring a delay, and maybe this will fix more other such cases.

with requiring a delay

You mean "without requiring a delay"?

Well, Safari doesn't support requestIdleCallback so I was testing with requestAnimationFrame and the problem still exists there with this method. Investigating further, turns out that without a delay the element is found and clicked too early -- requestIdleCallback was inherently causing a delay to occur so that in the end it worked.

So I am wondering now: should we click only elements which are visible? There is a method to check visibility, so the scriptlet could use this before clicking. Could this cause issues to existing trusted-click-element filters?

I added the when-visible: modifier for selectors: if a selector is prepended with when-visible:, the matching element will be clicked only once it becomes visible. gorhill/uBlock@aa028ea. Example:

voelkner.de##+js(trusted-click-element, 'when-visible:button[data-cookie_consent="0"]')

So this changes nothing for existing trusted-click-element filters.

should we click only elements which are visible

There are some cases where the sites hide the reject buttons but we can still click on them. Would this affect these cases?

I introduced the prefix when-visible: for when we want to wait for visibility, so all existing filters are unaffected by the addition of this prefix.

The biggest change from the latest commit is that the scriptlet no longer depends on mutation observer to wait for targeted elements to become available, hopefully no regression arise from this change.