I'm not even sure this needs a proposal. It seems like clearly just a bug.

Inclined to agree; removing "proposal" tag

Comment by Sergey Minakov (JIRA)

Is this bug present only on UnsafeMutableBufferPointer<Type> with range subscript as it's the only setter on marked with 'nonmutating' attribute? I would like to fix it, but don't quite get it yet.

Sorry naithar (JIRA User), I don't follow your question. Care to rephrase?

In case it helps, the reason for this bug is that despite its conformance to Collection, UnsafeMutable[Raw]BufferPointer has the semantics of a pointer; it does not own its elements, so mutating the elements should not be seen as a mutation of the buffer pointer.

Comment by Sergey Minakov (JIRA)

@dabrahams sorry 🙂
I've ment that after looking into source, it seems that only UnsafeMutableBufferPointer<T>.subscript(range: ) requires the value to be mutable. So changes should be made only to subscript(range: ) method. Did I undestand correctly?

naithar (JIRA User)Yes I think you understood correctly!

Comment by Sergey Minakov (JIRA)

@dabrahams is there any preferable way to implement this?
I have done changes that seem to work, but I'm not sure that's entirely correct as I had to change the validation tests: here is the commit.
Creating a temporary mutable value that would be used in `_writeBackMutableSlice` also doesn't seem like a good idea to me.

#12504 implements a nonmutating range setter for UnsafeMutableBufferPointer. The other setters are nonmutating at this point.

#12504 got merged; marking this as resolved.