[Lldb-commits] [lldb] Convert the StackFrameList mutex to a shared mutex. (PR #117252)

[via lldb-commits]

jimingham wrote:

Does this look better?

The situation is that there are really only three "writer" API's: `GetOnlyConcreteFramesUpTo`,` `Clear` and the second half of `GetFramesUpTo`, and only Clear is called with the intention of changing the list contents.  And none of these were client API's.  All the client API's were formally readers - they shouldn't have to know that we update frames lazily.  So the responsibilities were already pretty strictly limited.

But clearer is for sure better, so I renamed `GetOnlyConcreteFramesUpTo` to `FetchOnlyConcreteFramesUpTo` and moved the second half of GetFramesUpTo into `FetchFramesUpTo`.  Those and Clear are the only places where the list is actually mutated.  `Clear` doesn't need to do any dance, since it's a client API and intends to change the list.  Other than that, clients who want to read the list need to acquire the shared mutex, if the `Fetch` functions need to add frames to fulfill that request, that should happen silently when some client asks for more frames than have been fetched.

https://github.com/llvm/llvm-project/pull/117252