Previously the output of the SPIRVCross HLSL/MSL/WGSL pass were used as input to the reflection parsing, which required to create dummy samplers, which in turn could lead to issues like https://github.com/floooh/sokol-tools/issues/109.
The new approach uses common SPIRVCross GLSL output for the reflection parsing (which doesn't require to deal with dummy samplers and seems to produce the correct output for parsing the reflection data - also leading to simpler code).
Previously the output of the SPIRVCross HLSL/MSL/WGSL pass were used as input to the reflection parsing, which required to create dummy samplers, which in turn could lead to issues like https://github.com/floooh/sokol-tools/issues/109.
The new approach uses common SPIRVCross GLSL output for the reflection parsing (which doesn't require to deal with dummy samplers and seems to produce the correct output for parsing the reflection data - also leading to simpler code).