This feels like the same kind of issue mesa just had around the zlib update breaking downstream user programs (viewperf). If there are significant downstream issues for users you shouldn’t upgrade, even if that is the end goal.

Projects that are big and important get old and bloated because they need to try and span legacy issues alongside their attempts at newer paradigms. It’s just kind of the natural lifecycle of these projects.

I already force Wayland global for SDL games because the xwayland one has a horrible stutter while the native Wayland works flawlessly. Making it the default sounds reasonable to me. If specific programs don’t work with it, they can override it

Wayland is the future. It has already surpassed X11 in many ways. My favourite comment on Phoronix was “When is X11 getting HDR? I mean, it was released 40 years ago now.”

That said, the fact that this pull request came from Valve should carry some weight. Perhaps Wayland really is not ready for SDL.

I do not see why we need to break things unnecessarily as we transition. This is on the app side. Sticking with X11 for SDL ( for now ) does not harm the Wayland transition in any way. These applications will still work fine via Xwayland.

Sure, a major release like 3.0 seems like a good place to make the switch. In the end though, it is either ready or it is not. If the best path for SDL is to keep the default at X11 then so be it ( for now ).

one of the main reasons the linux kernel is where it is today: “never break userspace”

unfortunately not every project keeps to this principle.