On Fri, Jan 03, 2025 at 12:39:08PM +0100, Jakub Jelinek wrote:
> Should be all !DECL_NAME (decl) MK_unique? Or just
> DECL_DECOMPOSITION_P (decl) && !DECL_NAME (decl)?
> Structured bindings can't be redeclared, so I guess they are unique.
With
--- gcc/cp/module.cc.jj 2025-01-02 11:47:10.367499411 +0100
+++ gcc/cp/module.cc 2025-01-03 13:22:38.529485672 +0100
@@ -10985,6 +10985,12 @@ trees_out::get_merge_kind (tree decl, de
break;
}
+ if (DECL_DECOMPOSITION_P (decl))
+ {
+ mk = MK_unique;
+ break;
+ }
+
if (IDENTIFIER_ANON_P (DECL_NAME (decl)))
{
if (RECORD_OR_UNION_TYPE_P (ctx))
instead the second testcase doesn't ICE during -fmodule-header saving
anymore, but during import (on the simplified testcase):
0x11ba712 crash_signal
../../gcc/toplev.cc:322
0x891d9b is_bitfield_expr_with_lowered_type(tree_node const*)
../../gcc/cp/typeck.cc:2403
0x8920c0 is_bitfield_expr_with_lowered_type(tree_node const*)
../../gcc/cp/typeck.cc:2443
0x892105 is_bitfield_expr_with_lowered_type(tree_node const*)
../../gcc/cp/typeck.cc:2449
0x89215a unlowered_expr_type(tree_node const*)
../../gcc/cp/typeck.cc:2468
0x4270a2 build_new_op(op_location_t const&, tree_code, int, tree_node*,
tree_node*, tree_node*, tree_node*, tree_node**, int)
../../gcc/cp/call.cc:7153
0x89a2e1 build_x_binary_op(op_location_t const&, tree_code, tree_node*,
tree_code, tree_node*, tree_code, tree_node*, tree_node**, int)
../../gcc/cp/typeck.cc:4794
0x6d4f64 cp_parser_binary_expression
../../gcc/cp/parser.cc:10818
0x6d556c cp_parser_assignment_expression
../../gcc/cp/parser.cc:10979
0x6d5aa8 cp_parser_expression
../../gcc/cp/parser.cc:11162
0x6ddfe1 cp_parser_condition
../../gcc/cp/parser.cc:14308
Jakub
