[Bug middle-end/66214] [6 Regression] ICE verify_type failed with -O0 -g via gen_type_die_with_usage's dwarf2out.c:20250

["hubicka at gcc dot gnu.org" <gcc-bugzilla at gcc dot gnu dot org>]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=66214

Jan Hubicka <hubicka at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |jason at redhat dot com

--- Comment #8 from Jan Hubicka <hubicka at gcc dot gnu.org> ---
Jaon,
the issue here is that TYPE_CANONICAL is incomplete type:
 <record_type 0x3fff8e94c8d0 A readonly type_5 VOID
    align 8 symtab 0 alias set -1 canonical type 0x3fff8e94c8d0 context
<translation_unit_decl 0x3fff96fc0078 D.1>
    full-name "const class A"
    n_parents=0 use_template=0 interface-unknown reference_to_this
<reference_type 0x3fff8e94ca20>>

while the type itself is complete (it is a completetion of the class A provided
later):

 <record_type 0x3fff8e94d158 value_type readonly used type_5 VOID
    align 8 symtab -1903885616 alias set -1 canonical type 0x3fff8e94c8d0
    fields <field_decl 0x3fff8e834c00 m_upperBorder
        type <integer_type 0x3fff8e770690 int public type_6 SI
            size <integer_cst 0x3fff8e7316b0 constant 32>
            unit size <integer_cst 0x3fff8e7316c8 constant 4>
            align 32 symtab 0 alias set -1 canonical type 0x3fff8e770690
precision 32 min <integer_cst 0x3fff8e731668 -2147483648> max <integer_cst
0x3fff8e731680 2147483647>
            pointer_to_this <pointer_type 0x3fff8e7717a0>>
        private decl_3 VOID file t.C line 7 col 7
        align 1 offset_align 1 context <record_type 0x3fff8e94c4e0 A>
        chain <type_decl 0x3fff8e834b68 A type <record_type 0x3fff8e94ceb8 A>
            nonlocal decl_4 VOID file t.C line 6 col 9
            align 1 context <record_type 0x3fff8e94c4e0 A> result <record_type
0x3fff8e94c4e0 A>
           >> context <translation_unit_decl 0x3fff96fc0078 D.1>
    full-name "const value_type"
    n_parents=0 use_template=0 interface-unknown reference_to_this
<reference_type 0x3fff8e94d200>>

Shouldn't the canonical type be always the complete variant of the type?