[lldb/Gui] zero-initialize children_stop_id

[rupprecht]

Summary

This is currently causing msan warnings in the API tests when run under msan, e.g. commands/gui/basic/TestGuiBasic.py.

I'm not sure if 0 is the right initial value, but the tests pass with it (and msan doesn't complain anymore).

[clayborg]

lldb/source/Core/IOHandlerCursesGUI.cpp

1499–1511

Or we can inline the simple hard coded values into the definitions:

uint32_t children_stop_id = 0;
int row_idx = 0;
int x = 1;
int y = 1;
bool might_have_children = false;
bool expanded = false;
bool calculated_children = false;
std::vector<Row> children;


Row(const ValueObjectSP &v, Row *p)
    : value(v, lldb::eDynamicDontRunTarget, true), parent(p),
      might_have_children(v ? v->MightHaveChildren() : false) {}

[rupprecht]

SG, will land momentarily with the one formatting change. Thanks!

lldb/source/Core/IOHandlerCursesGUI.cpp

 
 } // namespace curses
 
 using namespace curses;
 
 struct Row {
   ValueObjectManager value;
   Row *parent;
   // The process stop ID when the children were calculated.
-  uint32_t children_stop_id;
-  int row_idx;
-  int x;
-  int y;
+  uint32_t children_stop_id = 0;
+  int row_idx = 0;
+  int x = 1;
+  int y = 1;
   bool might_have_children;
-  bool expanded;
-  bool calculated_children;
+  bool expanded = false;
+  bool calculated_children = false;
   std::vector<Row> children;
 
   Row(const ValueObjectSP &v, Row *p)
-      : value(v, lldb::eDynamicDontRunTarget, true), parent(p), row_idx(0),
-        x(1), y(1), might_have_children(v ? v->MightHaveChildren() : false),
-        expanded(false), calculated_children(false), children() {}
+      : value(v, lldb::eDynamicDontRunTarget, true), parent(p),
+        might_have_children(v ? v->MightHaveChildren() : false) {}
 
   size_t GetDepth() const {
     if (parent)
       return 1 + parent->GetDepth();
     return 0;
   }
 
   void Expand() { expanded = true; }