Don't possibly forget type definition when reading a CorpusGroup

Suppose type T was declared in the main corpus of a group, and suppose
it's a declaration-only type.  Later, when loading another corpus of
the group, we see a definition of T.  We should load the definition of
T and not just say that we have already seen T from the main corpus
and we just keep its declaration and never get its definition.

This is what this patch does.

	* src/abg-dwarf-reader.cc (add_or_update_class_type): Look for
	declaration-only-ness to determine if we've already seen the same
	type from the main corpus of the group.

Signed-off-by: Dodji Seketeli <dodji@redhat.com>

diff --git a/src/abg-dwarf-reader.cc b/src/abg-dwarf-reader.cc
index 0dd5c92b2599f6efd16cf2acf5f17a43477d1bb5..c95237c9da8e643904ef1bba764a5bd0ac98a7c6 100644 (file)
--- a/src/abg-dwarf-reader.cc
+++ b/src/abg-dwarf-reader.cc
@@ -12978,11 +12978,23 @@ add_or_update_class_type(read_context&         ctxt,
            // TODO: if there is just one class for that name defined,
            // then re-use it.  Otherwise, don't.
            result = lookup_class_type(name, *corp);
-         if (result)
+         if (result
+             // If we are seeing a declaration of a definition we
+             // already had, or if we are seing a type with the same
+             // declaration-only-ness that we had before, then keep
+             // the one we already had.
+             && (result->get_is_declaration_only() == is_declaration_only
+                 || (!result->get_is_declaration_only()
+                     && is_declaration_only)))
            {
              ctxt.associate_die_to_type(die, result, where_offset);
              return result;
            }
+         else
+           // We might be seeing the definition of a declaration we
+           // already had.  In that case, keep the definition and
+           // drop the declaration.
+           result.reset();
        }
     }