Re: [Boost-users] time_period constructor question

From: Ben Hutchings (ben.hutchings_at_[hidden])
Date: 2005-01-27 10:00:48

Jeff Garland wrote:
> On Wed, 26 Jan 2005 20:04:57 +0000, Ben Hutchings wrote
> 
>>Jeff Garland wrote:
>>
>>>Yeah, you are correct it is treating it as a function declaration.  This feel
>>>like a gcc bug to me, but there could be some obscure reason why it's that way
>>>that I'm unaware of.
>>
>><snip>
>>
>>This is correct behaviour.  The above declaration declares t2 as a 
>>function taking two parameters of type ptime (irrelevantly named 
>>min_date_time and max_date_time).  Any compiler that doesn't parse 
>>it as that is broken.
> 
> 
> So enlighten me.  Why would parenthesis be allowed in the function definition
> / name?

Parentheses are sometimes needed in the middle of declarators (example: 
int (*pf)()), and are allowed even if they are not needed (example: int 
(i)).  That applies even where the names will be ignored, as in the case 
of parameter names for a non-defining function declaration.  This 
surprising behaviour is sometimes called the "most vexing parse".

> Can't say I've ever seen that syntax used before.  What's the
> advantage -- I can see the downside.  

I wouldn't argue in favour of it.

Ben.