Fix T90170: `RNA_property_pointer_get` creating data in non-thread-safe way.

Protect this accessor with a local static mutex when it needs to
create/write data.

Ideally accessors should never create or modify data, but there are some
cases where this bad behavior is currently unavoidable.

This is the case of the Pointer accessor when the actual IDProperty has
not yet been created.

NOTE: this fixes a memory leak in liboverride diffing process when
several different overrides use a same linked reference ID.

Differential Revision: https://developer.blender.org/D12060

source/blender/makesrna/intern/rna_access.c

@@ -36,6 +36,7 @@
 #include "BLI_dynstr.h"
 #include "BLI_ghash.h"
 #include "BLI_math.h"
+#include "BLI_threads.h"
 #include "BLI_utildefines.h"
 
 #include "BLF_api.h"
@@ -3676,6 +3677,8 @@ PointerRNA RNA_property_pointer_get(PointerRNA *ptr, PropertyRNA *prop)
   PointerPropertyRNA *pprop = (PointerPropertyRNA *)prop;
   IDProperty *idprop;
 
+  static ThreadMutex lock = BLI_MUTEX_INITIALIZER;
+
   BLI_assert(RNA_property_type(prop) == PROP_POINTER);
 
   if ((idprop = rna_idproperty_check(&prop, ptr))) {
@@ -3695,9 +3698,14 @@ PointerRNA RNA_property_pointer_get(PointerRNA *ptr, PropertyRNA *prop)
     return pprop->get(ptr);
   }
   if (prop->flag & PROP_IDPROPERTY) {
-    /* XXX temporary hack to add it automatically, reading should
-     * never do any write ops, to ensure thread safety etc. */
+    /* NOTE: While creating/writing data in an accessor is really bad design-wise, this is
+     * currently very difficult to avoid in that case. So a global mutex is used to keep ensuring
+     * thread safety. */
+    BLI_mutex_lock(&lock);
+    /* NOTE: We do not need to check again for existence of the pointer after locking here, since
+     * this is also done in #RNA_property_pointer_add itself. */
     RNA_property_pointer_add(ptr, prop);
+    BLI_mutex_unlock(&lock);
     return RNA_property_pointer_get(ptr, prop);
   }
   return PointerRNA_NULL;