Actually if we want to improve this I would go for:

• std::vector
• reserve
• emplace_back

That combination would give us zero-copy initialization of the vector and can unfortunately not be achieved with Qt's container classes.

That's due to change in Qt6.
Given that's not too far away, I don't think we should be mixing two different container class types across the core. We're just making work for ourselves. Especially when this is all just premature optimisations.

In D18198#391797, @davidedmundson wrote:

That's due to change in Qt6.

Oh, I wasn't aware of that. Where do you have that information from? I'd love to read up about it.

Given that's not too far away, I don't think we should be mixing two different container class types across the core. We're just making work for ourselves. Especially when this is all just premature optimisations.

I doubt it's premature. The code is called every frame - small things count.

In D18198#391769, @graesslin wrote:

• reserve

Would it be okay?

phase2.reserve(stacking_order.size());

Or should we use some heuristic, e.g. num clients / num desktops?