Re: [PATCH 3 PR gdb/16959] gdb hangs in infinite recursion

[Wei-min Pan <weimin.pan@oracle.com>]

On 3/24/2018 5:33 AM, Simon Marchi wrote:
> On 2018-03-24 06:10, Pedro Alves wrote:
>> On 03/23/2018 09:25 PM, Weimin Pan wrote:
>>
>>> -  if (TYPE_CODE (type) == TYPE_CODE_STRUCT)
>>> +  struct type *real_type = check_typedef (type);
>>> +  if (TYPE_CODE (real_type) == TYPE_CODE_STRUCT)
>>>      {
>>>        CORE_ADDR *first_dont_print;
>>>        CORE_ADDR addr;
>>> @@ -658,15 +659,14 @@ cp_print_static_field (struct type *type,
>>>        addr = value_address (val);
>>>        obstack_grow (&dont_print_statmem_obstack, (char *) &addr,
>>>              sizeof (CORE_ADDR));
>>> -      type = check_typedef (type);
>>> -      cp_print_value_fields (type, value_enclosing_type (val),
>>> +      cp_print_value_fields (real_type, value_enclosing_type (val),
>>
>> This is still passing the resolved type down instead of the
>> original type.
>
> I did not point this out because cp_print_value_fields does a 
> check_typedef anyway, so it doesn't change anything.  But it's true 
> that to be consistent it would be better to always pass down the 
> original type, and let the callee decide whether it wants to resolve 
> the typedef or not.  Please consider this comment when preparing the 
> next version.
>

Please note the check_typedef() call (now redundant and removed) before 
calling cp_print_value_fields().
So passing the resolved type is correct.

Thanks,
Weimin

> Thanks,
>
> Simon