Re: std::ext STL (was: Re: design doc on alternative pointer support)

[Paolo Carlini <paolo dot carlini at oracle dot com>]

Hi,

> I just made STL list container work properly with 90% of my
> test application (shifted_ptr_test2.cpp) and tuned up the patch files I have
> sent last week. 
> They can be found here, for the interested:
> http://svn.boost.org/svn/boost/sandbox/shifted_ptr/bits/

Just wanted to make clear (again) some basic points:
1- If (when) you are going to submit for official review a patch of yours, you will require a Copyright assignment on file:

  http://gcc.gnu.org/contribute.html

That takes time, beware.
2- Related to the above, but not completely, at this moment I'm assuming you are going to propose something with Bob. To date, only Bob has a Copyright assigment on file, AFAIK. If not, at some point the two different approaches will have to be compared and contrasted.
3- To date, there are a few open options, listed below, for the general structure of the work. I was under the impression the first one was viable and favored. The third one seems also worth exploring:
- Code living in std::, special instantiations for "alternative pointer" types, along the way proposed by Bob some time ago. Bob told me privately he is working on a set of testcases (which always come with a new feature), but that will take him some more time.
- Code living in __gnu_cxx:: (/ext as a dir), completely decoupled from std:: code, a pure extension.
- Code living in a new namespace, an approach completely analogous to the debug-mode and parallel-mode facilities. Of course binary compatibility with normal containers cannot be achieved, but the user, at variance with the previous solution, is allowed to not change the sources, has simply to recompile all the translation units using the containers (the so called "Per-use recompilation" in the documentation of debug-mode)

Paolo.