[Lldb-commits] [lldb] [lldb] Recurse through DW_AT_signature when looking for attributes (PR #107241)

[Pavel Labath via lldb-commits]

================
@@ -60,44 +60,45 @@ class DWARFDebugInfoEntry {
     return attrs;
   }
 
-  dw_offset_t
-  GetAttributeValue(const DWARFUnit *cu, const dw_attr_t attr,
-                    DWARFFormValue &formValue,
-                    dw_offset_t *end_attr_offset_ptr = nullptr,
-                    bool check_specification_or_abstract_origin = false) const;
+  dw_offset_t GetAttributeValue(const DWARFUnit *cu, const dw_attr_t attr,
+                                DWARFFormValue &formValue,
+                                dw_offset_t *end_attr_offset_ptr = nullptr,
+                                bool check_elaborating_dies = false) const;
 
-  const char *GetAttributeValueAsString(
-      const DWARFUnit *cu, const dw_attr_t attr, const char *fail_value,
-      bool check_specification_or_abstract_origin = false) const;
+  const char *
+  GetAttributeValueAsString(const DWARFUnit *cu, const dw_attr_t attr,
+                            const char *fail_value,
+                            bool check_elaborating_dies = false) const;
 
-  uint64_t GetAttributeValueAsUnsigned(
-      const DWARFUnit *cu, const dw_attr_t attr, uint64_t fail_value,
-      bool check_specification_or_abstract_origin = false) const;
+  uint64_t
+  GetAttributeValueAsUnsigned(const DWARFUnit *cu, const dw_attr_t attr,
+                              uint64_t fail_value,
+                              bool check_elaborating_dies = false) const;
----------------
labath wrote:

Most of the time, I'd say. DW_AT_specification says "this DIE is an out of line definition of the function declared in <ref>". It normally does not list the name of the function because the name is already given on the other die, but it'd perfectly reasonable for `GetName()` on the first die to return the value from the other die. The same holds for most of the other attributes (with DW_AT_declaration and DW_AT_sibling being the most notable exceptions). I think this situation is similar enough to how DW_AT_signature works for types for it to get the same treatment.

https://github.com/llvm/llvm-project/pull/107241