> On Sep 2, 2016, at 2:41 PM, René J.V. Bertin <[email protected]> wrote:
> 
> On Friday September 02 2016 13:45:56 Lawrence Velázquez wrote:
> 
>>> Or is it something that will in fact mostly/only benefit Linux users?
>> 
>> Yes. This is meant to keep Clang compatible with ABI changes to GCC 5.1's 
>> libstdc++.
> 
> That's about time ... GCC 6.1 is out ...

My impression is that GCC's own implementation was unstable and buggy and 
difficult to track.

>> P.S. Incidentally, this implies that g++ 4.x is also ABI-incompatible with 
>> libstdc++ 5.1. Have we encountered any issues along these lines?
> 
> Funny enough I haven't noticed any issues with that on Linux

It's possible that MacPorts itself might not see any fallout from this because 
we've been strongly discouraging port maintainers from using g++ for quite a 
while.

> at some point I started using GCC 5 (and now GCC 6) without rebuilding my 
> whole system, and haven't run into any issues. Maybe that means that 
> libstdc++ is linked in as a private library and designed in such a way that 
> different versions can be loaded in memory?

Without doing any research, I can't say I know anything about how GCC is 
packaged on Linux, although we did previously end up with separate libstdc++ 
dylibs for each GCC port, until Jeremy hacked libgcc together.

It's also possible that 5.1 maintains backward compatibility but breaks forward 
compatibility.

vq
_______________________________________________
macports-dev mailing list
[email protected]
https://lists.macosforge.org/mailman/listinfo/macports-dev

Reply via email to