Updated examples to be compatible with self-hosted

[kcbanner]

• Uses the new asRange calling convention from https://github.com/floooh/sokol/pull/710
• Change Mat4/Vec2/Vec3 to extern struct as arrays in packed structs are no longer supported, and a guaranteed memory layout is needed for passing these to the GPU

All the examples (except for instancing.zig) run for me on self-hosted (Windows, x86_64) in both opengl and directx. I did notice the point size seems different between dx and gl for sgl-points, but this seems true with -fstage1 as well.

The instancing.zig example crashes on self-hosted with a stack overflow, at the start of frame(). Debugging this, it seems that it tries to pass ~40M to ___chkstk_ms, which comes from lib/compiler_rt/stack_probe.zig. I recall seeing some activity about this so this may just be a self-hosted bug. I only have access to Windows, so it may not affect other platforms. This may also be related to the issue with structs being copied on to the stack (https://github.com/ziglang/zig/issues/12638) but I'm not familiar with how the stack probe is supposed to work, so I can't say without more investigating.

[jeffkdev]

My own zig implementation of the sokol instancing example is is actually what triggered this issue: https://github.com/ziglang/zig/issues/12568

If you do a similar workaround to the one I posted there you can probably get the instancing to work on stage2 too.

[jeffkdev]

For reference, this is what my update loop looks like now on my side (I don't have the data wrapped in a state struct and just use the sokol c code directly, so maybe it won't work here):

    {
        var i: u32 = 0;
        // update particle positions
        while (i < cur_num_particles) : (i += 1) {
            const particle_pos =  &pos[i];
            const particle_vel =  &vel[i];

            particle_vel.y -= frame_time;
            particle_pos.* = particle_pos.add(particle_vel.scl(frame_time));
            // bounce back from 'ground'
            if (particle_pos.y < -2.0) {
                particle_vel.y = -1.8;
                particle_vel.y = -particle_vel.y;
                particle_vel.x *= 0.8;
                particle_vel.y *= 0.8;
                particle_vel.z *= 0.8;
            }
        }
    }

[kcbanner]

Ah, thanks! Indeed that workaround works here as well, just added the change. Cheers!

[floooh]

Ok, I have merged this PR manually into a new sokol-zig branch 0.10.0:

https://github.com/floooh/sokol-zig/tree/zig-0.10.0

...once Zig 0.10.0 is officially released, I'll merge that back into master, and (hopefully), this PR should then automatically resolve too.

PS: ...and I have converted this PR to a draft so that I don't accidentally press the merge button ...

PPS: many thanks for the PR :)

[floooh]

...and it's in master now.

[kcbanner]

awesome!