Whitespace changes behavior of folding @code ranges in fuse

[ryzngard]

In design time the parser will assign the trailing whitespace differently and ultimately causing the folding range to be different. Causing many of the CodeFoldingTests to fail due to span differences.

With Fuse

Without Fuse

[jjonescz]

Changing the runtime codegen would be a breaking change. Can we avoid that? What exactly does tooling need here?

[ryzngard]

The problem here is I'm not sure we in tooling can tell the difference. The structure that we're looking at doesn't have an indication of leading/trailing whitespace. The FullSpan and Span are the same, I don't see anywhere in a CSharpCodeBlock doesn't have a way to tell if it's ending in a blank line in the span or not. We'd have to do text parsing which seems to be against intent here. Maybe there's another way we can strip leading/trailing whitespace? In Roslyn I would just remove the trivia but I don't think it's so simple here.