Event input code requires two renders?

Nobody has claimed this yet.

So, I've encountered this problem specifically on the edit-box. It's in this code:

` if (in && in->mouse.buttons[NK_BUTTON_LEFT].clicked && in->mouse.buttons[NK_BUTTON_LEFT].down) {
edit->active = NK_INBOX(in->mouse.pos.x, in->mouse.pos.y,
bounds.x, bounds.y, bounds.w, bounds.h);
}`

Left-mouse-button must be clicked and currently held down for an edit-box to be made active (show the caret).

However, if in a single event pump I push both the left-mouse-button down and left-mouse-button up, ctx->input->mouse.buttons[0].down is 0 and ctx->input->mouse.buttons[0].clicked is 2, and the edit-box is never made active.

Speaking specifically about Windows, is there some guarantee that each cycle of the event loop will only receive once mouse button event? The demos appear to work, but my code is a little more complicated, and yet I think it's only a matter of timing that the demos work, and if the events were delivered a little more batch-like, the demos would also fail.

What am I missing here?

Start with the edit-box input handling shown in the issue and trace how the Windows event pump updates mouse button state when down and up arrive in one cycle. Reproduce the batched-event case and determine the expected activation behavior; done means the behavior is defined and the edit-box handles that event sequence consistently.

Written by the indexing model from the issue text.