What a convoluted mess.
At this point, it feels like a matter of time before Rust replaces C/C++.
Multiply that code base size by like, 78.3, and you’re possibly in the same galaxy as the all the c++ codebases out there that will be maintained for the next 50 years.
Rust may eat the lunch of c++ moving forward, the language will never go away.
Just like COBOL! Seriously, _just like COBOL_. The language will fade in importance over time. C++ will be relegated to increasingly narrow niches. It will eventually only be found in "legacy" systems that no one wants to spend to rewrite. No one young will bother at all to learn C++ at all. Then, in a few decades, they'll be pulling folks like you and me out of retirement to maintain old systems for big bucks.
If the new language has good interop with the old then this doesn’t have to be painful either. A while back, I was working on a large Objective-C code base which was slowly moving to Swift. We quickly got to a point where new code was written in Swift by default. In only a couple of years, we had a code base which used Objective-C pretty much only in the dark corners where code churns slowly.
"In Rust that's just pushing the problem to the borrow checker and codegen which has plenty of memory corruption issues too but just calls it "bugs". In C++ you avoid those issues by reducing dynamic memory allocations to a minimum, and using checked APIs as 'safety boundaries' instead of directly poking around random arrays. Different approach but not any less safe than C++".
Both statements are pretty ridiculous. It's pretty clear that moving up in terms of the safe abstractions that are possible makes your code safer because the safety is there when you reach for that kind of programming paradigm & the programming paradigms you need are typically a property of the problem domain rather than the language (it's it's the intersection of "how is this problem solved" and "what tools does the language give you"). In C it gives you few tools and you either twist into a pretzel trying to stick to that or you write our own tools from scratch every time and make the same but different mistakes over and over. No language is perfect and all will guide you to different takes on the same problem that better suit to its paradigm, but there are intractable parts that will force you to do things (e.g. heap allocation and more involved ownership semantics are not uncommon). Moreover C very limited ability to manage your own codebase to define API boundaries - the only tool is opaque types with a defined set of functions declared in 2 different files.
The fact that it's mostly backwards compatible means you can reproduce almost all issues of c in c++ awell, but the average case fares much better. Real world C++ does not have double-frees, for example. (As real world C++ does not have free() calls).
> it's safer and higher productivity than C or assembly
Debatable - in C and even more so in assembly, the unsafety is at least directly in your face, while in C++ it's cleverly diguised under C++ stdlib interfaces (like dangling std::string_views, all the iterator invalidation scenarios etc... meaning that memory corruption bugs may be much harder to investigate in C++ than in C (or even assembly) - which in turn affects productivty (e.g. what's the point in writing higher level buggy code faster when debugging takes much longer).
> it feels like a matter of time before Rust replaces C/C++
It may replace C++, but C will most likely outlive at least C++, and I wouldn't be surprised if Rust too - even if just as C API wrappers to allow code written in different modern languages to talk to each other ;)
I've never seen any users who avoid using Rust's stdlib on principle. The closest thing is the use of the parking_lot crate over the standard Mutex types, but that's getting better over time, not worse, as the stdlib moves closer to parking_lot in both implementation and API. Other than that, there's only been one "wrong" API that needed to be deprecated with a vengeance (mem::uninitialized, replaced by mem::MaybeUninit). Especially compared to C++, the fact that Rust doesn't have a frozen ABI means it's free to improve the stdlib in ways that C++ can only dream of. While I do wish that the Rust stdlib contained some more things (e.g. a standard interface to system entropy, rather than making me pull in the getrandom crate), for what it provides the Rust stdlib is quite good.
Care to provide examples?
I think they generally do a very good job at curating the APIs to feel very consistent. And with editions, they can fix almost any mistakes they make whereas c++ doesn’t have that. In fact, the situation in c++ is so bad that they can’t fix any number of issues in basic containers and refuse to introduce updated fixes because “it’s confusing”. The things they keep adding to the stdlib aren’t even core useful things that come out of the box. Like missing a networking API in 2025. The reason is they have to get it perfect because otherwise they can’t fix issues and they can’t get it perfect because networking is a continuously moving target. Indeed we managed to get a new tcp standard before c++ even though it had even worse ossification issues. Or not caring about the ecosystem enough to define a single dependency management system because the vendors can’t get agreement or a usable module system. Or macros being a hot mess in c++ some 20 years after it was already known to really suck.
Now it’s possible given enough time rust will acquire inconsistent warts over time similar to c++, but I just don’t think it’ll ever be as bad in the standard library due to editions and automated migration tools being easier against rust. Similarly, I think editions give them the flexibility to address even many language level issues. Inertia behind ways of doing things are the harder challenges that are shared between languages (eg adoption of Unsend might prevent the simpler Move trait from being explored and migrated to), but c++ is getting buried under its own weight because it’s not a maintainable language in terms of the standard because the stewards refuse to realize they don’t have the necessary editorial tools and keep not building themselves such tools (even namespace versions ended up being aborted and largely unused).
https://doc.rust-lang.org/std/option/enum.Option.html
Same with Result, Box, Rc, Arc, Cell, RefCell, (plus even more *Cells), Iter, ... the resulting 'bread crumbs syntax' is IMHO the main reason why typical Rust code is so hard to read. This philosophy to push what should be language features into the stdlib is the same as in C++ (and IMHO one of the main problems of C++), it's probably also a reason why both languages are so slow to compile.
I expect Rust's successor might have a shot at replacing C++.
By the time C++ was as old as Rust, it had conquered the world. If Rust coulda, it woulda.
As soon as you start writing big(ger) software in Rust, its lacking ergonomics really become apparent.
.as_mut().unwrap().unwrap().as_mut().do_stuff() gets really old after a while.
And I am not convinced that borrow checking is the panacea that it's made out to be. Region-based memory management can accomplish the same safety goals (provided there's a sane concurrency model), without forcing people into manually opting into boxing and/or reference counting on a per-reference basis.
Throw into that the pain of manual lifetime management (it's not always elided, and when it needs to change, it's painful), I honestly believe it's far more reasonable to ask programmers to throw shit into two or three arenas and then go hog-wild with references than it is to expect them to deal with the tediousness of the way Rust does things.
We are just cargo-culters (no pun intended).