Re: [Boost-docs] Doxygen features

Subject: Re: [Boost-docs] Doxygen features
From: Steven Watanabe (watanabesj_at_[hidden])
Date: 2010-06-01 01:34:56

AMDG

Daniel James wrote:
> I was assuming that 'type' couldn't be a function, only a function
> pointer. But of course, it's used in typedefs which can be functions.
>   

type can be a function, since
void f(int());
is equivalent to
void f(int(*)());
and doxygen doesn't normalize the
function declarations.

> So... we'd need to skip over anything in angle or curly brackets and
> then distinguish between function pointers and normal functions.
> Tricky, when you consider (hopefully, these are all correct, I've
> never got the hang of member pointers):
>
> void(int*) => void f(int*)
> void(*)() => void(*f)()
> void(foo::*)() => void (foo::*f)()
>   

We would also have to deal with cv qualifiers for member pointers.

> I guess changing boostbook is probably the best route.
>   

I'm not sure that it has to be perfect.  Doxygen
generates gibberish for some of the complicated
tests I tried, anyway.

> btw. I just noticed that my patch breaks function definitions. So
> don't use it, I'll write something better soon.
>   

In Christ,
Steven Watanabe