4 ms·

C++ is an insanely versatile language, and since you can use any C library with "extern C", it has a huge amount of high-quality libraries available for it. ffm

C++ is an insanely versatile language, and since you can use any C library with "extern C", it has a huge amount of high-quality libraries available for it. ffmpeg, openSSL, webkit, Qt. the problem is there is not a cross platform package manager to manage these libraries, and manually downloading and compiling libraries is a hugely painful and error-prone process. it's so bad that big C++ projects like Firefox and Chromium basically write their own scripts that download and build dependencies, and simply commit the source code of the libraries they use directly into their own tree. I believe there is some thought being put into making a C++ package standard [0], which would do wonders for the language. in my opinion half of the reason Node took off is because npm is so easy to use, and people are getting excited about Rust partially because Cargo is so well thought out. Now I might be wrong here, but I think if you took modern C++ (no raw pointer manipulation) and combined it with a good package manager, I think you would pretty much erase any advantages Rust offers. [0]: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2016/p0235r0.pdf http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2016/p023...

fsloth 10y ago
"manually downloading and compiling libraries is a hugely painful and error-prone process." I think you are exaggerating. Most projects with any user base provide ready binaries which able to get going pretty fast. "it's so bad that big C++ projects like Firefox and Chromium basically write their own scripts that download and build dependencies, and simply commit the source code of the libraries they use directly into their own tree." I would not call this 'bad' but resilient. "Replicating is bad" is a non value adding rule of thumb, imo. That said, yes, it's very hard to start incorporating third party libraries to C++ when compared to any other language.
- gcp 10y ago
  Most projects with any user base provide ready binaries which able to get going pretty fast. But what about the dependencies of those binaries/libraries? Manually playing that cat and mouse game just sucks.
- TillE 10y ago
  Every single C++ project of any significance I've worked on or seen in the wild has some awful dependency compilation script/process. It sucks. There are no good cross-platform solutions. Even if everybody is using CMake perfectly, you still have to glue all the pieces together. But hey, modules are coming in the next couple years. Should really help with this.
gcp 10y ago
I think if you took modern C++ (no raw pointer manipulation) and combined it with a good package manager, I think you would pretty much erase any advantages Rust offers. I've seen this opinion now twice in short succession on HN. Issues with C++ safety aren't just a matter of a dangling pointer that should've been a unique_ptr or shared_ptr. It's massively naive to think that this would in any way be comparable to Rust. That's not even considering the reality that no real-life C++ codebase seems to survive on those modern C++ features alone. (If it did, I presume Mozilla would've rewritten Firefox in C++11/14/17 instead of Rust. Oh wait, they already use C++11...)
- duneroadrunner 10y ago
  > Issues with C++ safety aren't just a matter of a dangling pointer that should've been a unique_ptr or shared_ptr. It's massively naive to think that this would in any way be comparable to Rust. While I agree, I would point out that in the same way that "safe" Rust is a usable (if difficult to get used to) subset of "unsafe" Rust, there also exist usable safe subsets of C++ [1]. Tooling to enforce that Rust code strictly conform to the safe subset, of course, already exists (and is built into the compiler). Such enforcement tools do not yet exist for C++ code, but I think they're coming. And I think they'll probably support safe subsets that are more convenient/intuitive/easier-to-use than (the current version of) safe Rust. I wonder if adopting Rust to avoid the pervasive unsafe code in C++ isn't a little bit akin to adopting Esperanto to avoid the pervasive cursing in English. I mean, there does exist a non-vulgar subset of the English language, even if it's difficult to persuade people to stick to it :) But then, if all the cool kids are talking Esperanto... > (If it did, I presume Mozilla would've rewritten Firefox in C++11/14/17 instead of Rust. Oh wait, they already use C++11...) I think this may be a missed opportunity by the Firefox developers. If you look at the code, they're still using new/delete and raw pointers targeting dynamic objects. I presume that a lot the code has survived from the pre-C++11 era. I think converting their code to a safe subset of C++ (like SaferCPlusPlus) would be quicker and require less effort than a rewrite in Rust. [1] for example (shameless plug): https://github.com/duneroadrunner/SaferCPlusPlus https://github.com/duneroadrunner/SaferCPlusPlus
  - none_for_me_thx 10y ago
cousin_it 10y ago
Does this one-liner count as modern C++? const string& s = max<string>("a", "b"); It makes a dangling reference that will crash when accessed. In Rust that's impossible.
- pjmlp 10y ago
  Nice example.
- symlinkk 10y ago
  good point, I had never heard of that.

fsloth 10y ago

"manually downloading and compiling libraries is a hugely painful and error-prone process." I think you are exaggerating. Most projects with any user base provide ready binaries which able to get going pretty fast. "it's so bad that big C++ projects like Firefox and Chromium basically write their own scripts that download and build dependencies, and simply commit the source code of the libraries they use directly into their own tree." I would not call this 'bad' but resilient. "Replicating is bad" is a non value adding rule of thumb, imo. That said, yes, it's very hard to start incorporating third party libraries to C++ when compared to any other language.

gcp 10y ago
Most projects with any user base provide ready binaries which able to get going pretty fast. But what about the dependencies of those binaries/libraries? Manually playing that cat and mouse game just sucks.
TillE 10y ago
Every single C++ project of any significance I've worked on or seen in the wild has some awful dependency compilation script/process. It sucks. There are no good cross-platform solutions. Even if everybody is using CMake perfectly, you still have to glue all the pieces together. But hey, modules are coming in the next couple years. Should really help with this.

gcp 10y ago

I think if you took modern C++ (no raw pointer manipulation) and combined it with a good package manager, I think you would pretty much erase any advantages Rust offers. I've seen this opinion now twice in short succession on HN. Issues with C++ safety aren't just a matter of a dangling pointer that should've been a unique_ptr or shared_ptr. It's massively naive to think that this would in any way be comparable to Rust. That's not even considering the reality that no real-life C++ codebase seems to survive on those modern C++ features alone. (If it did, I presume Mozilla would've rewritten Firefox in C++11/14/17 instead of Rust. Oh wait, they already use C++11...)

duneroadrunner 10y ago
> Issues with C++ safety aren't just a matter of a dangling pointer that should've been a unique_ptr or shared_ptr. It's massively naive to think that this would in any way be comparable to Rust. While I agree, I would point out that in the same way that "safe" Rust is a usable (if difficult to get used to) subset of "unsafe" Rust, there also exist usable safe subsets of C++ [1]. Tooling to enforce that Rust code strictly conform to the safe subset, of course, already exists (and is built into the compiler). Such enforcement tools do not yet exist for C++ code, but I think they're coming. And I think they'll probably support safe subsets that are more convenient/intuitive/easier-to-use than (the current version of) safe Rust. I wonder if adopting Rust to avoid the pervasive unsafe code in C++ isn't a little bit akin to adopting Esperanto to avoid the pervasive cursing in English. I mean, there does exist a non-vulgar subset of the English language, even if it's difficult to persuade people to stick to it :) But then, if all the cool kids are talking Esperanto... > (If it did, I presume Mozilla would've rewritten Firefox in C++11/14/17 instead of Rust. Oh wait, they already use C++11...) I think this may be a missed opportunity by the Firefox developers. If you look at the code, they're still using new/delete and raw pointers targeting dynamic objects. I presume that a lot the code has survived from the pre-C++11 era. I think converting their code to a safe subset of C++ (like SaferCPlusPlus) would be quicker and require less effort than a rewrite in Rust. [1] for example (shameless plug): https://github.com/duneroadrunner/SaferCPlusPlus https://github.com/duneroadrunner/SaferCPlusPlus
- none_for_me_thx 10y ago

cousin_it 10y ago

Does this one-liner count as modern C++? const string& s = max<string>("a", "b"); It makes a dangling reference that will crash when accessed. In Rust that's impossible.

pjmlp 10y ago
Nice example.
symlinkk 10y ago
good point, I had never heard of that.