Manuel Teira wrote:
Yes, yes. I'm sorry for feeling the forum with short posts.
Don't be! Its been very useful and informative.
I will try to collect more information to minimize the number of posts. However, after reading that document, I've reached the conclusion that without a binder2nd specialization for mem_fun_ref, this is not going to work (perhaps the gnu STL is providing such a specialization). What I've made, since the boost::bind is already used in some parts of that source code, is to migrate every bind2nd + mem_fun_ref combination to a boost::bind approach, resulting in the following patch:
I came to the same conclusion, i.e. use boost bind throughout to avoid the problem (and for consistency).
I've checked in a change that I think is pretty much identical to your patch. Thanks for persevering and helping us improve things!
