[Lldb-commits] [lldb] [lldb-dap] Ensure we acquire the SB API lock while handling requests. (PR #137026)

[Jonas Devlieghere via lldb-commits]

JDevlieghere wrote:

> High level, I do not think it is a good idea to hold the top level API mutex which is way too large locking scope. We should leave the decision to each request handler for smaller scope locks if they want to ensure a critical section. Can we revert the PR?

FWIW I don't necessarily agree with this observation. While I agree that a smaller critical section is better, conceptually a single request corresponds to a group of SB API calls. Making it the responsibility of the requests to group all SB API calls within the scope of the lock is a lot more error prone.

What is the other lock involved in the deadlock, other than the API mutex?

https://github.com/llvm/llvm-project/pull/137026