[Lldb-commits] [lldb] [lldb] Fix crash after second run when set a previous watchpoint. (PR #136682)

[via lldb-commits]

llvmbot wrote:


<!--LLVM PR SUMMARY COMMENT-->

@llvm/pr-subscribers-lldb

Author: None (hapeeeeee)

<details>
<summary>Changes</summary>

This PR fixes a crash in `LLDB` caused by a dangling pointer to a reused `ValueObjectSP` when re-running the debuggee and setting the same watchpoint again.

As described by @<!-- -->jasonmolenda, the fix is to reinitialize the dangling pointer in `Watchpoint::SetEnabled`.

This PR closes [#<!-- -->135590](https://github.com/llvm/llvm-project/issues/135590).

---
Full diff: https://github.com/llvm/llvm-project/pull/136682.diff


1 Files Affected:

- (modified) lldb/source/Breakpoint/Watchpoint.cpp (+9) 


``````````diff
diff --git a/lldb/source/Breakpoint/Watchpoint.cpp b/lldb/source/Breakpoint/Watchpoint.cpp
index 2df848aaa0576..0fcc9b90c0ab5 100644
--- a/lldb/source/Breakpoint/Watchpoint.cpp
+++ b/lldb/source/Breakpoint/Watchpoint.cpp
@@ -409,6 +409,15 @@ bool Watchpoint::IsDisabledDuringEphemeralMode() {
 }
 
 void Watchpoint::SetEnabled(bool enabled, bool notify) {
+  // Whenever setting the enabled state of a watchpoint, we need to ensure
+  // that `m_new_value_sp` exists to avoid crash when reading old_data later.
+  // See https://github.com/llvm/llvm-project/issues/135590.
+  if (!m_new_value_sp) {
+    ExecutionContext exe_ctx;
+    m_target.GetProcessSP()->CalculateExecutionContext(exe_ctx);
+    CaptureWatchedValue(exe_ctx);
+  }
+
   if (!enabled) {
     if (m_is_ephemeral)
       ++m_disabled_count;

``````````

</details>


https://github.com/llvm/llvm-project/pull/136682