David Neto

You've convinced me. I also looked up Java's use of 'require'. > 7.7.1 Dependencies > The requires directive specifies the name of a module on which the current module has...

In the above it looks like only stage names were agreed to be shortened. That landed long ago. What remains is shortening other things like builtin and interpolation. Removing 'editorial'...

Discussed in the 2023-10-10 meeting. Decided to not revisit now.

Sorry for not keeping up on this. The group has agreed to do the following: - for const-expr and override-expr, anything generating a NaN or infinity generates an error at...

#3508 introduced indeterminate values and gave general rules. The rest is editorial, to propagate and describe the cases for each builtin functions.

@alan-baker rewrote the variable and value sections in [#3226 ](https://github.com/gpuweb/gpuweb/pull/3226), and clarified that initializers for `private` address space variables must be a const-expression or an override-expression.

@alan-baker and I have been iterating on a subgroups proposal internally to Google. There's a lot of overlap with this proposal (including the same spelling for a lot of builtins)....

I understand that the Vulkan CTS has a design principle that the CTS framework has to know whether it can run a test before it gets too deep into looking...

> But it looks like the "fail" case of validateRequirements is treated as an error rather than "don't run me" Oh, looks like that's only tried in a verify coherency...

Vulkan test examination occurs here, in vkt::TestCaseExecutor::init https://github.com/KhronosGroup/VK-GL-CTS/blob/main/external/vulkancts/modules/vulkan/vktTestPackage.cpp#L417 vktCase->checkSupport(*m_context); vktCase->delayedInit(); m_progCollection.clear(); vktCase->initPrograms(sourceProgs); The constraint is that checkSupport is a const method, and to say a test should not be run,...