David Neto

spec change https://github.com/gpuweb/gpuweb/pull/4981

I only got as far as using the "easy build" flow. https://github.com/google/filament/blob/main/BUILDING.md#easy-build

I made a fresh docker image based on Ubuntu Noble 24.04, installed cmake, ninja-build, gcc, and clang. Then specifying the CC and CXX and CXXFLAGS as suggested also fails: ```shell...

When I modify the command line to use -libstdc++ I get past configuration but fail compilation because it can't find the C++ headers. CC=/usr/bin/clang CXX=/usr/bin/clang++ CXXFLAGS=-stdlib=libstdc++ ./build.sh -c release |...

This one will take some care in reviewing. I *think* it's all I meant to fix up for "infinity" in milestone 0. (CR)

For my own sake: The resolution is to add an fine-grain (expression level?) diagnostic to allow const-eval'd INF and NaN to be valid, and implementations replace them with 0.

> GLSLANG_BUILD_SHARED_LIBS is itself an anti-pattern. I'm coming to that belief too. People have said that of Shaderc before, and they've won me over. Related: - supporting shared libs "properly"...

Refreshing for myself: * WGSL currently doesn't concern itself whether filtering is occuring during texture operations. * the rules apply to (texture,sampler) _pairs_ that appear in sampling operations (with prefix...

If I swap the last two rows and the last two columns of the compatibility table in the proposal, I get this: | | Sampler `non_filtering` | Sampler `unknown` |...

Hm. I guess the checks weren't as horribly coupled as I feared. There's nice monotonicity here. You have to check for the properties of the *worst* sampler that a texture...