Do not parse methods with accessibility modifiers as local functions (alt)

[CyrusNajmabadi]

Fixes #74482

Alternate to #74484

Also, makes the parser much better about errant close curly tokens causing type-members to now show up within a namespace, instead of in the preceding type-declaration.

[CyrusNajmabadi]

Currently draft. But pulling @333fred in to get thoughts.

[CyrusNajmabadi]

Added as the internal syntax generator doesn't include these abstract/overrides in that model.

[CyrusNajmabadi]

i really don't ove this. but this mirrors how normal public syntax works. And, importantly, means that if we ever add a new type-decl subclass, we'll know we have to ensuer that thsi works.

[RikkiGibson]

It looks like some editor features tests are failing

[CyrusNajmabadi]

It looks like some editor features tests are failing

Both cases where hte new behavior in the presence of these error cases is fine :)

[CyrusNajmabadi]

@RikkiGibson this is ready for review.

[RikkiGibson]

I am a little uneasy about this diagnostic. There is nothing wrong in itself with putting a } here. The problem is that the } is followed by a non-type/non-namespace member declaration.

I would be more comfortable with a diagnostic which blames the member private void M() while implying that we didn't expect the preceding }.
Something like: A namespace cannot directly contain members such as fields, methods or statements. Did you mean to include it as a member of type 'Type'? (similar message to ERR_NamespaceUnexpected)

It's possible that I just need to sit with it a bit and I'll come around. I'm not sure.

[CyrusNajmabadi]

this is fair. i'll think on it too and check in with you later.

[CyrusNajmabadi]

having thought about it, i'm ok with this error message. it is invalid. Now it happens to be invalid because of what we see later, btu i'm ok with that.

note: i'm not pushing back strongly. if you do want a specific error here, lmk.

[CyrusNajmabadi]

A namespace cannot directly contain members such as fields, methods or statements.

i'm not a fan of this primarily because the tree we produce won't jive with this. The node will appear inside the class from all other perspectives. But i do think we can clarify the error we put on the }. Something like Errant '}' found between type members. By saying 'between' we call out that there is another member tha tfollows.