This brings render pass handling closer to Metal / WebGPU:
sg_pass_action now has separate store actions
sg_pass now has separate resolve-target images
filled some gaps in sg_begin_pass() re image types
msaa images no longer have an implicit resolve-texture, instead this needs to be a separate sg_image
more detailed validation in sg_make_image, sg_begin_pass and sg_apply_bindings
TODO:
[x] rename sg_image_desc.render_attachment back to sg_image_desc.render_target (Metal calls the flag 'render target' while WebGPU calls it 'render attachment', we'll just stick with Metal in this case)
[x] ...also in samples
[x] fix d3d11 samples
[x] dummy backend
[x] fix tests
[x] add tests for sg_make_image validation checks
[x] add tests for sg_make_pass validation checks
[x] update documentation comments
[x] update readme and changelog
[x] fix language bindings samples
[x] investigate and fix the Rust bindings clear-color bug (the clear sample background color is green-cyan instead of yellow-red) => nvm, the C library just wasn't recompiled, deleting the target directory fixed the problem (still strange, looks like cargo's dirty-checks don't work for compiling C libraries?)
This brings render pass handling closer to Metal / WebGPU:
TODO:
sg_image_desc.render_attachmentback tosg_image_desc.render_target(Metal calls the flag 'render target' while WebGPU calls it 'render attachment', we'll just stick with Metal in this case)investigate and fix the Rust bindings clear-color bug (the clear sample background color is green-cyan instead of yellow-red) => nvm, the C library just wasn't recompiled, deleting the(still strange, looks like cargo's dirty-checks don't work for compiling C libraries?)targetdirectory fixed the problem