Supporting C++11 on old OS X (Was: kdevelop 4.7 and beyond (was: clang++-mp-3.4 doesn't find initializer_list on OS X 10.6))

[Lawrence Velázquez]

On Sep 8, 2014, at 2:08 PM, Brandon Allbery <allbery.b at gmail.com> wrote:

> On Mon, Sep 8, 2014 at 1:24 PM, Lawrence Velázquez <larryv at macports.org> wrote:
>> That's true, but it's a red herring. If our libstdc++ wasn't broken, ports would not have to avoid it.
>> 
>> Our actual problem is that none of the g++ compilers we distribute are viable because they don't use the system's C++ runtime. If we successfully move the whole operation over to libc++, they'd still be broken.
> 
> I think you have misunderstood the problem completely, unless you intend to conclude that everything before Mavericks must never even attempt to support C++11.

I apologize if I haven't been clear. I've been mixing two tangentially related topics. I'll try again.

- My first assertion is that MacPorts' libstdc++ is broken on OS X because it doesn't use the system C++ runtime, libc++abi. This assertion, strictly speaking, is not about C++11 support.

- My second assertion is that if MacPorts' libstdc++ were not broken in this way, ports would be able to use it for C++11 support on older OS X, if they wanted. We wouldn't have to mass-migrate everything to libc++.

My points hinge on an admittedly tenuous assumption that the critical issue is the C++ runtime (libc++abi vs. libsupc++) and not the C++ standard library (libc++ vs. libstdc++). Clarification on this point would be welcome.

It also occurs to me just now that Snow Leopard probably doesn't have libc++abi, but I can't check at the moment. Clarification on this would also be welcome.

vq