Mikko Mononen on Nostr: nprofile1q…xjvvf That makes sense. I had another look at the nanovg out-of-atlas ...

nprofile1qy2hwumn8ghj7un9d3shjtnddaehgu3wwp6kyqpqqw9huvn8gh2svc3r7vlc9qcwy2l8rg0yxfzht4pd5e0z4s9j3jtqhxjvvf (nprofile…jvvf) That makes sense. I had another look at the nanovg out-of-atlas case and looks like it just creates a bigger atlas (doubles the shorter side), but keeps the current atlas around until the end of the frame, so no need to patch any data.

https://github.com/memononen/nanovg/blob/master/src/nanovg.c#L2418

I've experimented with some quite elaborate resizing/defragging in the past, but the complex ones always got out of hand. I think the most useful one was a version that kept the rasterized data around, but would do full relayout.