[Bug middle-end/61558] [5 Regression] ICE: Segmentation fault

[hubicka at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61558

--- Comment #11 from Jan Hubicka <hubicka at gcc dot gnu.org> ---
It seems that C++ FE do not produce any assembler name for b (because it is not
instantiated?).
(gdb) p debug_tree (node->decl)
 <var_decl 0x7ffff6ae4bd0 b
    type <integer_type 0x7ffff6adb690 int public type_6 SI
        size <integer_cst 0x7ffff6af9048 constant 32>
        unit size <integer_cst 0x7ffff6af9060 constant 4>
        align 32 symtab 0 alias set -1 canonical type 0x7ffff6adb690 precision
32 min <integer_cst 0x7ffff6af9000 -2147483648> max <integer_cst 0x7ffff6af9018
2147483647>
        pointer_to_this <pointer_type 0x7ffff6afd738>>
    public private static tree_0 external tls-local-dynamic decl_3 decl_6 SI
file bug.cc line 4 col 23 size <integer_cst 0x7ffff6af9048 32> unit size
<integer_cst 0x7ffff6af9060 4>
    align 32 context <record_type 0x7ffff6c472a0 A>
    template-info 0x7ffff6c443a0 chain <type_decl 0x7ffff6c354c0 A>>


In such case b should never land the symbol table.

I am looking into why the node is created.