[high-level] Should we support aliasing predeclared enumerant values? (e.g. user-defined @location(pos) mapping to @location(position))

[kdashg]

From #4505, related to #4539.

[kdashg]

WGSL 2024-03-26 Minutes

• KG: Seems kind of nice we can make an alias for, but M3+. Seems niche.
• DN: M3+ or later.
• AB: The discussion isn’t about when, but rather if the design features will fit together well.
• KG: Think it has limited use.

[kdashg]

WGSL 2024-04-16 Minutes

• (AB/DN: This is actually coupled to [high-level] Should predeclared enumerant values (e.g. position) should be shadowable by user variables? (e.g. let position =) #4539: Deciding that those attributes are not shadowable, means we can’t really make aliases for these things. That’s because defining and using an alias requires identifier resolution to be operating, and that means we get back into the world where we’re going to accidentally shadow position again.)
  • DN: In later internal discussion, it was pointed out you could solve this by putting position in its own name resolution space.
• KG: In M3+, was a concern about it last time. If we wanted to do this, we could, but punt for now, yes?
• DN: Yes.
• KG: Resolved, M3+.