Clarify Valid Values language in the attributes table.

[dj2]

Currently the Valid Values column says things like integer literal, yielding a positive i32 which I took to mean it could be 1u as an integer literal and the compiler would treat it as i32. This seems to be incorrect and this means it can be AbstractInt or i32 but not u32.

We should make explicit what types can be provided to the various attribute values. Even just making a list like:

• AbstractInt
• i32
• Must be positive

Could be clearer?

[dneto0]

That three-bullet list isn't quite right because its "or" between the first two, and "and" with the last.

Before the last change I had suggested

integer literal evaluating to a positive i32 value

Would that be any clearer?

The 1u cannot evaluate to an i32 value because there is no feasible automatic conversion

[dneto0]

Also, we've deliberately decided to keep the compilers simple by keeping these as literals, not creation-time expressions.
See #2790 (post-V1) and its duplicate #2890

So we definitely want the wording that it must be a literal.

[dj2]

Even just:

• AbstractInt or i32 literal
• Must be positive

the evaluating could work but spelling out the types is more explicit.

[dneto0]

Works for me.

[dneto0]

Currently says things like

Must be a positive i32 or AbstractInt literal

It's not clear how far the "or" distributes.

What would be better is to write two sentences: one about types, the other about value.

Must be i32 or AbstractInt
Must be positive.