[Bug c++/93667] [10 regression] ICE in esra with nested [[no_unique_address]] field

[jamborm at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=93667

--- Comment #5 from Martin Jambor <jamborm at gcc dot gnu.org> ---
It is easy to prevent the ICE with the following, which prevents total
scalarization from happening.  However, if someone marked a field with
such an attribute, the encompassing structure perhaps should be
optimized a much as we can?

So I am thinking of adding a predicate
bunch_of_empty_records_p which will return true for a type which only
consists of records which only have field_decls which are other
records but nothing else and still allow total scalarization of
those.

The easy fix is:

--- a/gcc/tree-sra.c
+++ b/gcc/tree-sra.c
@@ -958,6 +958,9 @@ scalarizable_type_p (tree type, bool const_decl)
   if (type_contains_placeholder_p (type))
     return false;

+  bool have_predecesor_field = false;
+  HOST_WIDE_INT prev_pos = 0;
+
   switch (TREE_CODE (type))
   {
   case RECORD_TYPE:
@@ -966,6 +969,14 @@ scalarizable_type_p (tree type, bool const_decl)
        {
          tree ft = TREE_TYPE (fld);

+         HOST_WIDE_INT pos = int_bit_position (fld);
+         if (have_predecesor_field
+             && pos <= prev_pos)
+           return false;
+
+         have_predecesor_field = true;
+         prev_pos = pos;
+
          if (DECL_BIT_FIELD (fld))
            return false;