Reflection should be at the top or at least static compile time reflection should be there. It opens doors to so many other things and makes others much easier but I believe it should be a compiler thing and not an STL thing...
"interface()" looks interesting, but would it also help with debug / release builds as those seem to generate problems just as often when deploying binaries to users? While that addition looks to fix a lot of problems, I'm still quite inclined towards the "eww" option (C).
I always wondered why the special math functions are in the standard. Sure, they are in the C standard, but why? Almost nobody uses them and only a few people in the world know how to implement them.
Thanks for the clarifications on ABI and I think that brainstorming about solutions could give us some good options. One Question, Why is mixing ABIs on the same compute unit important ? Could Modules be used to version the STL ? Won't Modularisation of the STL be an opportunity to do a clean ABI break ?
I think that the comment about c++11 only now reaching everyone is telling. How much effort is being duplicated among all the different compilers in maintain their libraries? What is even the point of talks on C++23 features when no one can use them until 2025-6,and get won't reach full penetrative for another 10 years? We need a more python-like upgrade schedule.
Great talk although I felt that inline namespaces issues exposed were not totally fair : if an author of a struct X containing a std::string must provide ABI stability to its clients, then putting X too into its dedicated, versioned and tagged inline namespace looks like the simple solution too me.
OK, what are the "few" (more than one) other planets that have devices running C++ on them? I can only suppose some Mars probes use C++, but what other planets even have running devices from Earth on them?
Personally I think you wrong about what makes C++ successful. I think it’s the emotional energy people such as yourself use in your work is what makes c++ successful. I feel like that’s what draws me to the language personally.
Why does each vendor's library implementation need to implement each feature on its own, from scratch? Why can't they copy each other or work as a group?
![Don't constexpr All the Things - David Sankel [CppNow 2021]](http://i.ytimg.com/vi/NNU6cbG96M4/mqdefault.jpg)
![Don't constexpr All the Things - David Sankel [CppNow 2021]](/img/tr.png)
![Iterators and Ranges: Comparing C++ to D to Rust - Barry Revzin - [CppNow 2021]](http://i.ytimg.com/vi/d3qY4dZ2r4w/mqdefault.jpg)
![Algorithms from a Compiler Developer's Toolbox - Gábor Horváth - [CppNow 2021]](http://i.ytimg.com/vi/eeS1WP7FK-A/mqdefault.jpg)
![The C++ Rvalue Lifetime Disaster - Arno Schödl - [CppNow 2021]](/img/1.gif)
phenomenal content CppNow. I smashed the thumbs up on your video. Maintain up the terrific work.
Very very interesting. This talk must have taken a lot of work
Very good talk. Touched upon all the major challenges that std faces.
Reflection should be at the top or at least static compile time reflection should be there. It opens doors to so many other things and makes others much easier but I believe it should be a compiler thing and not an STL thing...
I'd imagine it requires both. Nice interfaces for th reflection in the STL. The reflection itself would have to be vendor implemented, I'd imagine.
"interface()" looks interesting, but would it also help with debug / release builds as those seem to generate problems just as often when deploying binaries to users?
While that addition looks to fix a lot of problems, I'm still quite inclined towards the "eww" option (C).
I always wondered why the special math functions are in the standard. Sure, they are in the C standard, but why? Almost nobody uses them and only a few people in the world know how to implement them.
Thanks for the clarifications on ABI and I think that brainstorming about solutions could give us some good options. One Question, Why is mixing ABIs on the same compute unit important ? Could Modules be used to version the STL ? Won't Modularisation of the STL be an opportunity to do a clean ABI break ?
I think that the comment about c++11 only now reaching everyone is telling. How much effort is being duplicated among all the different compilers in maintain their libraries? What is even the point of talks on C++23 features when no one can use them until 2025-6,and get won't reach full penetrative for another 10 years? We need a more python-like upgrade schedule.
Great talk although I felt that inline namespaces issues exposed were not totally fair : if an author of a struct X containing a std::string must provide ABI stability to its clients, then putting X too into its dedicated, versioned and tagged inline namespace looks like the simple solution too me.
OK, what are the "few" (more than one) other planets that have devices running C++ on them? I can only suppose some Mars probes use C++, but what other planets even have running devices from Earth on them?
Personally I think you wrong about what makes C++ successful. I think it’s the emotional energy people such as yourself use in your work is what makes c++ successful. I feel like that’s what draws me to the language personally.
”Just use cmake” 😰
Why does each vendor's library implementation need to implement each feature on its own, from scratch? Why can't they copy each other or work as a group?
it's *not* zero cost