[cfe-dev] on libc++ extensions and slist

[Christopher Jefferson]

On 31 Jul 2011, at 02:31, Howard Hinnant wrote:

> In existing C++03 standard libraries there are a ton of extensions.  Some good, some not so good.  Some of the better ones have morphed into official  C++11 libraries.  However, I can't think of a single example where an extension has been standardized unchanged from its extension form.  There is good reason for this.
> 
> I'm going to pick on slist.  But my argument is meant to be broader than slist.
> …
> 
I have converted some fairly substantial code bases at my work to run on clang, then libc++, then c++0x. I also did quite a lot of the work involved in getting boost to run on clang/libc++/c++0x.

On any sizeable code base, each of these changes, particularly mapping to clang and then libc++, will take some work. This work is mostly because other compilers have been far too lax in the past.

I want to see libc++ and clang++ keep their neatness, and not start introducing huge number of patches to get around old standards-incompatable code.

This certainly seems to be the route that clang++ itself is sticking to, my experience is almost any piece of sizeable c++ code will fail to compile when first given to clang++ (although the error messages are very helpful in showing how to fix it).

However, a written migration strategy, covering issues which are likely to come up, would be useful. I'm not sure what the best way to present such a document would be (in svn, on a wiki for easier editing?)

In short: Tell people how to fix their code, don't help them patch around it silently. They are going to have pain switching to clang/libc++/0x whatever you do.

Chris