[Lldb-commits] [PATCH] D157159: [lldb] Properly protect the Communication class with reader/writer lock

[Alex Langford via Phabricator via lldb-commits]

bulbazord added inline comments.


================
Comment at: lldb/source/Core/Communication.cpp:41-43
   Clear();
 
+  std::unique_lock guard(m_shared_mutex);
----------------
I think this one //may// be slightly incorrect. Here's how I imagine it going wrong:

- Thread 1 calls Connect. It only runs `Clear()` (which disconnects) and then Thread 2 is scheduled.
- Thread 2 then calls Connect. It also only runs `Clear()` and then Thread 1 is scheduled.
- Thread 1 grabs the lock, connects, and exits.
- Thread 2 also grabs the lock, connects, and exits.

I'm not sure how gracefully `Connection::Connect` handles another connection if not properly disconnected from the first url beforehand. Reading through some of the `Connection` derived class implementations of `Connect` I'm not sure exactly what will happen though. The documentation isn't much clearer either...


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D157159/new/

https://reviews.llvm.org/D157159