These are symptoms of language issues.
https://blogs.msdn.microsoft.com/vcblog/2015/12/03/c-modules...
Regarding templates, the approach is similar to the deprecated export keyword, so partially compiled templates and inline functions are stored in the metadata database used by the compiler for modules.
Bjarne was of course aware of modules, being brought on Simula and other programmer friendly languages, but then he had to make C++ code fit into UNIX linkers that only knew about AT&T Assembly and C binary formats.
So no modules there, and along the years you got a culture of developers, specially those that only learn on the job, that never learned about modules and its history.
You see it happen also when cache friendly code and RAII get discussed.
Those things were also possible outside the C family of languages, before they won the market. So most millennials think that they are some special language capabilities only possible in C and C++.
To be honest, I can't talk intelligently about cache-friendliness at all. I know that there exist models of computation (say, for complexity analysis purposes) that explicitly take memory hierarchies into account, but I've never seen them actually used on anything but the simplest data structures and algorithms.
With respect to RAII, I think the main benefit isn't cache-friendliness, but rather deterministic destruction - you can't delay relinquishing a scarce resource that someone else might want to use.
> So most millennials think that they are some special language capabilities only possible in C and C++.
I'm a millennial, and have to admit with shame that I grew up thinking C++ is the best thing ever. But it's possible to recover from it. :-)
Take a look at a typical performance-oriented GPU coding, cache is the major analysis parameter there. Data structures are mostly 2D and 1D arrays, but they may be scrambled in a very complex way, and the algorithms are arbitrarily complex.
I never said it was related to it, I said it was the other feature many think it is C++ specific, or introduced via C++.
Ada also has RAII via controlled types and Object Pascal also had destructors when it was introduced.
Functional programming languages that allow for rich macros, support trailing lambdas or currying also allow for RAII like implementations using those features.
For example, with-open-file in Common Lisp.
We can't say that (yet) about Swift and Rust. In 3-5 years I think this is going to be a very different conversation, though.
In my mind, C++'s flexibility is both its greatest benefit and its greatest danger. You can do almost anything, and there are so many ways to do it. I agree this is a "problem" that's likely not going to be fixed. You can ask your fellow developers to read Scott Meyer's Effective (Modern) C++, you can go to meetups, listen to the wisdom of the steering committee, etc. but at the end of the day it really boils down to the fact that your team needs to be committed to being resilient and responsible. That's true in any language, but much more so in C++.
Not sure about Swift, I haven't really been paying attention in that space.
Even then, in C++ you still need to expose a C interface, so it feels at least on even footing with Rust. But of course, I am biased...
Writing proper code in C++ requires, as you say, "your team needs to be committed to being resilient and responsible", which just doesn't happen on my little piece of the world.
So using it in personal projects, yeah. At work, not really.