https://gcc.gnu.org/bugzilla/show_bug.cgi?id=88112
--- Comment #4 from Eric Botcazou <ebotcazou at gcc dot gnu.org> --- > Index: gcc/tree.c > =================================================================== > --- gcc/tree.c (revision 266308) > +++ gcc/tree.c (working copy) > @@ -5260,7 +5260,7 @@ free_lang_data_in_one_sizepos (tree *exp > Note this should only happen for abstract copies so setting sizes > to NULL is OK (but we cannot easily assert this). */ > else if (expr && !is_gimple_val (expr)) > - *expr_p = NULL_TREE; > + *expr_p = build0 (PLACEHOLDER_EXPR, TREE_TYPE (expr)); > } > > > fixes the ICE but not sure if that's good. Setting to size_zero_node > also works (so much for "proper" type checking ...). Maybe it's better > to adjust the type-checker somehow? The C++ front-end should be fixed instead, i.e. it should gimplify all the size expressions (like the Ada front-end does).