Checks around ref & ref readonly required.

[Nigel-Ecma]

Adding this as an issue so it’s checked off before v7.

Recent issues have brought to light that the introduction of ref and ref readonly into v7 is unfortunately incomplete and where they are allowed and the rules governing them need to be checked to ensure v7 makes sense in this area.

Recent issues, and PRs if applicable, concerning ref & ref readonly:

• C# 7.3 method_body etc. grammar does not allow => ref #825, missing expression bodied functions variants (assigned, PR in progress)
• C# 7.3 interface_method_declaration etc. grammar does not allow return by reference #826, support in interfaces (assigned, not yet in progress)
• C# 7 §13.6.2 ref readonly in local_variable_initializer #828, initializers – addressed by PR ref readonly is not allowed in local_variable_initializer #831 (approved, pending merge)
• C# 7.3 §15.10.1 operator with in parameter modifier #835, in parameters for operators (self assigned, not yet in progress)

Issues not covered by PRs:

• Rules over assigning to ref & ref readonly variables should be checked for completeness (C# 7 §13.6.2 ref readonly in local_variable_initializer #828 suggests that some source material may have been sourced from earlier feature proposals and may not reflect the final intended semantics).
• Similarly rules for combining ref and ref readonly, e.g. when a selection expression has ref & ref readonly choices, need to be checked.

ref, a little word with a big impact!

The above list may not be exhaustive, please add to it if needed!

Of the existing PRs only one is “in progress” right now, next other two are in my queue – feel free to dive in and take an reassign, but maybe consider checking the rules for validity as that is not covered by any PR (and of course may produce none!)

[Nigel-Ecma]

Issue #894 has highlighted some more issues not mentioned above regarding read-only, e.g. passing a variable as a ref parameter does not specify that the variable cannot be read-only (note in is similar to ref readonly)

[jskeet]

Meeting 2023-09-06: Closing in favour of the specific examples in #894; there may well still be some things we've missed, but ultimately that will always be the case.