Tracking issue for code generation

[kainino0x]

For discussing assorted things related to the code generator (#259).

[kainino0x]

May 9 meeting:

.yml differences with dawn.json (FYI)

• AE: other day I decided to try to make dawn use the yaml file: convert from the yaml to our json file.
  • .yml doesn’t have default values
  • reserve 0-value of enums - either unnamed or “Undefined”. yml and dawn both are inconsistent right now
    • RM: Enum values are implicit for most enums in the yaml. There’s one or so that needed special numbering, so that is supported. WGPUSType_RenderPassDescriptorMaxDrawCount
    • Can put explicit numbers on all the enums if we need.
    • AE: Or be able to tell it to start at 1?
    • KN: Or do C-style numbering thing where you can set the first to 1, then it increments after that. But any solution is good
  • CreateRenderPipelineAsync with callback, assumes callback is called CreateRenderPipelineAsyncCallback, whereas dawn names them manually. Results in one difference: GetCompilationInfoCallback vs. CompilationInfoCallback
  • how to handle extra data - have an “extras: “ field in the .yml?
    • such as jsrepr (JS representation for Emscripten)
    • extras can be string->value map
    • and we can define extras in a separate yaml file
    • Should Go program be able to do the merging of multiple yamls into one json? Would require Go to be in the build process of Dawn so it can do the merging. Or just have each consumer do the merging it needs?