If I understand correctly, when the "validate" feature is not enabled, we should only do enough work to produce a ModuleInfo the backends can use, and no more. But naga::valid::Validator::validate_entry_point calls VaryingContext::validate on the entry point's arguments and return value, and I don't see that that contributes anything to the final ModuleInfo.
Should those calls be protected by a #[cfg(not(feature = "validate"))] condition?
If I understand correctly, when the
"validate"feature is not enabled, we should only do enough work to produce aModuleInfothe backends can use, and no more. Butnaga::valid::Validator::validate_entry_pointcallsVaryingContext::validateon the entry point's arguments and return value, and I don't see that that contributes anything to the finalModuleInfo.Should those calls be protected by a
#[cfg(not(feature = "validate"))]condition?