Re: [patch c,c++,i386]:PR/15774 - Conflicting function decls not diagnosed

[Kai Tietz <ktietz70 at googlemail dot com>]

2010/12/22 Kai Tietz <ktietz70@googlemail.com>:
> 2010/12/22 Kai Tietz <ktietz70@googlemail.com>:
>> 2010/12/22 Jason Merrill <jason@redhat.com>:
>>> On 12/22/2010 11:31 AM, Joseph S. Myers wrote:
>>>>
>>>> For compatibility testing, targetm.comp_type_attributes is the right thing
>>>> to use; no new hook should be needed. ?For printing, I don't see the need
>>>> for a new hook either. ?Why is it appropriate to print only attributes
>>>> that pass the new hook? ?Shouldn't you rather be printing all attributes,
>>>> regardless of whether they affect the calling convention, since they are
>>>> all part of the type in GNU terms?
>>>
>>> I'm inclined to agree.
>>>
>>> Jason
>>>
>>
>> Well, I talked yesterday to Ian about that, and I thought about adding
>> the attribute print even for tree-pretty-printer. Ian's said, that he
>> would accept here the printing of calling-convention attributes, but
>> he was strict against displaying all type attributes. So I added a
>> this hook for checking this, as obviously the comp_type_attributes
>> hook isn't usable for making this decission. Of course, if everybody
>> agrees, I can change the patch so that it displays all attributes.
>>
>> Regards,
>> Kai
>
> Well, I notice that I did a pasto and used wrong PR number. Sorry for
> that. It is http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12171 I meant.
>

I am working on that, too.  But this (PR/15774) problem is nothing for
stage 3 phase AFAICS. It would need to handle the diagnostic of the
comp_type_attributes-hook more strictly and add the missing places for
diagnostics.

Sorry,
Kai