On Thu, Apr 5, 2018 at 10:07 AM, Paolo Carlini <paolo.carl...@oracle.com> wrote:
> Hi,
>
> On 05/04/2018 15:56, Jason Merrill wrote:
>>
>> On Thu, Apr 5, 2018 at 8:27 AM, Paolo Carlini <paolo.carl...@oracle.com>
>> wrote:
>>>
>>> Hi,
>>>
>>> the main issue is already fixed in trunk but we still ICE on the reduced
>>> testcase attached by Jakub which has a broken std::initializer_list
>>> missing
>>> the definition. I think we can handle this case similarly to the existing
>>> check in do_pushtag, which would be also consistent with the plain error
>>> we
>>> give for, eg:
>>>
>>> namespace std { template <class> class initializer_list; }
>>>
>>> template class std::initializer_list<int>;
>>>
>>> However, we still have the option of issuing a fatal_error, like we do in
>>> finish_struct.
>>
>> How about using complete_type_or_maybe_complain instead of a custom error?
>
> Yes, I was about to send a message about that option, I already had it in
> the audit trail, tested too, then started fiddling with fatal_errors and
> forgot to mention it ;) Anyway, would be the below.OK, thanks. Jason
