Migrates to v0.14

[nicoabie]

@Interrupt I'm learning zig and I found delve very nice so wanted to contribute

This is missing emscripten for lua natecraddock/ziglua#86
I'll ask again if they can add that flag so we don't need to use your fork anymore}

for changes in the shaders
https://floooh.github.io/2024/11/04/sokol-fall-2024-update.html

[nicoabie]

addSharedLibrary is deprecated

[nicoabie]

new zig v0.14 didn't like that variable and function having the same name

[nicoabie]

this needs to be fixed to have the same funcitonality as with v0.13

[nicoabie]

I need to check this

[nicoabie]

new zig v0.14 didn't like that variable and function having the same name

[nicoabie]

I believe this is correct looking at https://github.com/floooh/sokol/pull/1111/files#diff-5a4303cccd1516d570a9b7a605d2b5ce0487743f686ec7a5fef129b88efa8500

[nicoabie]

breaks findLibraryFunctions

[Interrupt]

Is the bindings-cleanup branch explicitly required, hasn't this been merged in?

[nicoabie]

it's the latest tag. I though it made sense to link to the specific version of the bin which generated the shaders.

checked with latest master https://github.com/floooh/sokol-tools-bin/tree/9b5a3e2b57fe9783ba4d1f3249059bc4720b592f and the output is the same

[nii236]

Hey @nicoabie I had a quick test of your fork and none of the examples run correctly when compiled for WASM. I'm brand new to zig so not sure how to fix:

[nicoabie]

Hi! Thanks for reporting, I'll try to check it out in the following weeks. I'm new to zig as well :)

[Interrupt]

Sorry for the delay! Just digging into this now again.

When running the stresstest, something seems to be leaking vertex buffer bindslots. Getting the following error.

...
Skinned mesh example module initializing
Assertion failed: (a_state->buffer_index < SG_MAX_VERTEXBUFFER_BI

[Interrupt]

Still tracking this down, but it looks like if I stop caching the common pipelines in the platform/graphics.zig shader init functions, then the passes and stresstest run correctly:

    pub fn initFromBuiltin(cfg: ShaderConfig, comptime builtin: anytype) !Shader {
        const shader = try ShaderImpl.initFromBuiltin(cfg, builtin);
        // shader.makeCommonPipelines();
        return shader;
    }

[Interrupt]

Aha, I think it was a use-after-free in makeCommonPipelines - this seems to have fixed it:

    pub fn makeCommonPipelines(self: *Shader) void {
        for (common_vertex_layouts) |l| {
            _ = self.impl.makePipeline(l);
        }
    }