I Only Set the Same Value Again. The Icon Kept Flickering.

Call the Cat: Delivery — launch devlog

TL;DR: A shortcut button’s icon kept flickering. The image file was fine and the path was fine. The refresh code was assigning the same image path again on every redraw — and assigning the same value is not free.

I thought the image was the problem

The icon on the bottom-right shortcut button kept flickering. When an image request failed, it stayed as a broken-image placeholder.

I suspected the image file and the network first. The file was intact. The path was right.

The cause was the refresh code. Every time the screen redrew, it was writing the same path into the icon again.

Assigning the same value is still work

In a browser, assigning an image source re-runs the whole image loading procedure even when the address is identical to what was already there.

It feels like nothing should happen — the value is the same. It does not work that way. It requests again and paints again. For a brief moment in between there is an empty state, and that is what the eye reads as a flicker. When the request fails, the empty state stays and you get a broken image.

What made it bad is how often refreshes happen. The screen redraws whenever state changes, and the icon was being re-fetched every single time. In a game, state changes constantly, so the flickering never stopped.

The fix is simple. Assign only when the path has actually changed — one small function, and identical values do nothing. Elements that get rebuilt from scratch each time, like list rows, are not the target. Only long-lived elements need this.

Reassigning an identical address versus assigning only when it changed.

What happens when an AI writes “redraw”

Ask an AI to “refresh the screen when state changes” and it picks the most certain method: redraw everything.

That is not wrong code. The screen is always current and nothing is ever missed. The tests pass. Judged by the resulting values, it is perfect.

But what a person sees is not the resulting value — it is the act of redrawing. The cost of writing the same value again, the empty gap in between, the placeholder left behind on failure: tests do not look at any of that.

That is what happened here. The code behaved as intended and the tests passed. The flicker was only found when a human looked at the screen. Ask an AI “does it work?” and it answers pass. Ask “how does it look?” and it cannot answer at all.

One more thing surfaced in the same session. The script that checks release readiness kept reporting an already-solved item as blocked. The setting lived in a different env file, and the readiness script was only reading the default one. The checker and the actual configuration were looking at different places. A checker’s pass or fail is not something to take at face value either — look once at what it is reading.

Same value, refresh frequency, and the one tests cannot see — did you look at it.

Copy this — three lines for redraw code

When you get “redraw” code from an AI, look at these three.

Give the AI this one sentence before asking for screen refresh code.

Only apply a change to the screen when the value actually differs. If it is the same,
do nothing — and tell me which elements are rebuilt each time and which ones stay alive.

That sentence stops “redraw everything” from becoming the default.

At a glance

Solo dev of Call the Cat, shipped after 178 days
on Google Play
and Steam.
Now building the sequel, Call the Cat: Galaxy, in public.