[Lldb-commits] [lldb] [lldb] Change lldb's breakpoint handling behavior (PR #96260)

[Pavel Labath via lldb-commits]

================
@@ -377,6 +382,19 @@ class Thread : public std::enable_shared_from_this<Thread>,
 
   virtual void SetQueueLibdispatchQueueAddress(lldb::addr_t dispatch_queue_t) {}
 
+  /// When a thread has executed/trapped a breakpoint, set the address of that
+  /// breakpoint so we know it has been hit already, and should be silently
+  /// stepped past on resume.
+  void SetThreadHitBreakpointAtAddr(lldb::addr_t pc) { m_hit_bp_at_addr = pc; }
+
+  /// When a thread stops at a breakpoint instruction/address, but has not yet
+  /// executed/triggered it, record that so we can detect when a user adds a
+  /// breakpoint (or changes a thread to a breakpoint site) and we need to
+  /// silently step past that when resuming.
+  void SetThreadStoppedAtBreakpointSite(lldb::addr_t pc) {
+    m_bpsite_at_stop_pc = pc;
+  }
----------------
labath wrote:

I'm also finding it hard to wrap my head around the meaning of this variable. If I understand correctly it tells us: the pc that we've stopped at; that there was a breakpoint site there at the time of the stop; and we did not hit that site.

I'm wondering if it would be clearer if we unpacked that. What if we:
- unconditionally stored the PC of the last stop. Maybe this could be even a part of the StopInfo class, as I think it could be useful to see the PC value at the time of the stop, even if the user changed the PC afterwards.
- a flag indicating whether we stopped at a breakpoint site, regardless of whether we've hit it or not (per the previous comment, that could be indicated by the stop reason). This doesn't really look like it belongs to StopInfo class, but I think I'd be fine with putting it there for collocation purposes.

https://github.com/llvm/llvm-project/pull/96260