From: Peter Dimov (pdimov_at_[hidden])
Date: 2002-11-01 08:25:27
From: "David Abrahams" <dave_at_[hidden]>
> I'm currently trying to port Boost.Bind to work on Sun compilers, and
> I'm finding the structure of the code extremely difficult. Several
> factors contribute, including:
> 1. lack of comments
> 2. the use of cryptic names (which I presume are designed to help
> avoid symbol-length limitations)
> 3. The use of preprocessor macros not defined in the same file
> where they're used. Not a bad thing in itself, but it adds up...
The code started out pretty readable, but slowly evolved into its current
state, despite my efforts to keep it sensible. :-)
The main reasons for the complexity are (1) support for different calling
conventions, and (2) void returns workaround.
You might find older versions of bind.hpp much more easier to
None of the versions worked on Sun, however, IIRC. At one time the compiler
started to compile and run bind_test.cpp, but real world code still didn't
mem_fn_test, on the other hand, worked. Did it break?
> If somebody (Peter) could write a short guide to the implementation it
> might make things go much more smoothly. In particular it seems to be
> having trouble dealing with mem_fn(some-data-member-pointer). I have
> so far been unable to track down where member data pointers are
> distinguished from member function pointers!
Member pointers have a separate overload at mem_fn.hpp(320). Member function
pointer overloads are in mem_fn_cc.hpp.
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk