Re: [boost] Re: mem_fn warnings

From: Peter Dimov (pdimov_at_[hidden])
Date: 2004-03-14 06:12:13

David Abrahams wrote:
> "Peter Dimov" <pdimov_at_[hidden]> writes:
>
>> David Abrahams wrote:
>>> I keep getting lots of these:
>>>
>>> c:\boost\boost\mem_fn.hpp(264) : warning C4180: qualifier applied to
>>> function type has no meaning; ignored
>>>
>>> Would it be possible for the mem_fn header to #pragma push a warning
>>> disabler for MSVC?
>>
>> These warnings are probably a sign that the data member mem_fn/bind
>> overload is being picked instead of the correct member function
>> overload. They usually indicate a bug in your code. For example,
>> you've used bind(&X::f, _1) but &X::f doesn't take zero arguments.
>
> Wow, so they are!  I was getting other errors, and when I fixed those
> the warnings disappeared too.
>
> Incidentally, the errors I get when I misuse bind somehow (which
> happens frequently!) are *really* bad at telling me what I've done
> wrong.  Any chance of improving those?

They are, no question about it. It's mostly a question of supporting MSVC
6-class compilers, and partially a question of dependencies (type_traits,
for example).

I suspect I can do much better in a hypothetical 100% TR1 bind that only
targets 100% conforming compilers (and can freely use other TR1 libraries).
But what would the point be? Isn't Lambda supposed to fill that role? ;-)