Nested functions and tail-call optimization

[Ludovic Courtès]

Hi, Ian,

Ian Lance Taylor <iant@google.com> skribis:

> On Mon, Aug 20, 2012 at 2:25 AM, Ludovic Courtès <ludo@gnu.org> wrote:
>>
>> Consider this function:
>>
>>   static tree filter (bool (*pred) (const_tree), tree);
>>
>>   static tree
>>   list_remove (bool (*pred) (const_tree), tree t)
>>   {
>>     bool opposite (const_tree t)
>>     {
>>       return !pred (t);
>>     }
>>
>>     return filter (opposite, t);
>>   }
>>
>> Here, the call to ‘filter’ is likely subject to tail-call optimization.
>> Is there any guarantee that ‘opposite’ is still reachable when ‘filter’
>> is called?
>
> It would clearly be a bug for GCC to do a sibling call from
> list_remove to filter.

OK.

> I don't know whether GCC has that bug or not.

I haven’t witnessed it, just wondering.

Thanks,
Ludo’.