gltf.parse assumes 4-byte aligned little endian data

[fabioarnold]

Ran into this because I used @embedFile to embed my glb model. Unfortunately, according to ziglang/zig#4680 it's currently not possible to specify alignment for @embedFile.

I know it's simpler to just assume these preconditions (and it's probably faster to cast to u32 pointers). But it would be nice to at least have an assertion for alignment and endianness to help with the error search.

The safer/correct but also slower way would be to treat the data as a byte stream and use something like this:

var reader = std.io.fixedBufferStream(data).reader();
const my_int = reader.readInt(u32, .little);

Edit: nvm, saw in the glTF spec that 4-byte alignment is a requirement. So at least an assertion would be nice.

[kooparse]

Thanks, I will see what I can do on this; I'll let you know.

[fabioarnold]

Hey, thanks for considering this. And thanks for the great library. It's awesome!

I added align(4) to relevant places here: #10
Now type-checking will complain when you want to pass unaligned data to gltf.parse. What do you think?

I don't know if it's worth it to handle big endian. I think it's a pretty rare these days. I'm targeting wasm, btw.

[kooparse]

Merged; thanks a lot! :)