[Lldb-commits] [lldb] [lldb] Give thread specific async interrupt requests a lifetime (PR #224272)

[Raphael Isemann via lldb-commits]

================
@@ -4460,7 +4465,25 @@ Process::RunPrivateStateThread(PrivateStateThread::Purpose purpose) {
 
       continue;
     } else if (event_sp->GetType() == eBroadcastBitInterrupt) {
-      if (GetPublicState() == eStateAttaching) {
+      // Claim the request this event was sent for. A thread specific request
+      // (i.e. one coming from a thread plan rather than from the user) is only
+      // meaningful for the execution that was in flight when it was made: if
+      // the process has stopped since then, whatever the requestor wanted to
+      // interrupt is over. Interrupting the next resume instead would stop the
+      // process for a reason no thread plan can explain, which surfaces to the
+      // user as a spurious stop.
+      auto [interrupt_tid, stop_id] =
+          std::exchange(*m_pending_interrupt.Lock(), {});
----------------
Teemperor wrote:

We should have a `Pop` function or similar to make this less ugly.

https://github.com/llvm/llvm-project/pull/224272