Using C++

[powerboat9]

Describe the cleanup

Using some C++11 internally

It should be done because

Applying some C++ might cut down on boilerplate

What are the alternatives?

Not using C++

Additional context

I'm not quite sure what this would look like yet -- this is intended as more of a RFC to check if the answer is "absolutely not" before I try to put together a proof of concept.

Extra fields

• I have checked the existing issues
• I have read the Contributing Guidelines
• I'd like to work on this issue

[sskras]

I am unrelated to X11 development (yet).
But do you have any real life cases of "cutting down on boilerplate" code, please?
It would be nice to compare.

[shamefulCake1]

I am also unrelated to X11 development yet, but "some" c++ would be great, as long as the ABI remains in pure C, and there is a decent tooling to prevent contributors from going rampage with C++ magic. I am not sure whether such tooling exists, but something clang-tidy-related might be an option.

But do you have any real life cases of "cutting down on boilerplate" code, please?

I suggest enabling C++ features one-by-one, instead of "enabling C++".

For a start, I would enable just one (single) feature: RAII.
Very often we end up writing code like

void fun() {
int* a = NULL;
size_t of_a;
of_a = do_something_to_find_size();
...
a = (int*) calloc( of_a, sizeof(int));
... 
do_something_with_a( a );
free(a);
}

it is super easy to forget free(a);.

being able to use std::unique_ptr or a custom class which would auto-deallocate at the function end would be just indispensable.

[Sniperq2]

I am also unrelated to X11 development yet, but "some" c++ would be great, as long as the ABI remains in pure C, and there is a decent tooling to prevent contributors from going rampage with C++ magic. I am not sure whether such tooling exists, but something clang-tidy-related might be an option.

But do you have any real life cases of "cutting down on boilerplate" code, please?

I suggest enabling C++ features one-by-one, instead of "enabling C++".

For a start, I would enable just one (single) feature: RAII.
Very often we end up writing code like

void fun() {
int* a = NULL;
size_t of_a;
of_a = do_something_to_find_size();
...
a = (int*) calloc( of_a, sizeof(int));
... 
do_something_with_a( a );
free(a);
}

it is super easy to forget free(a);.

being able to use std::unique_ptr or a custom class which would auto-deallocate at the function end would be just indispensable.

But what's wrong with attribute((cleanup(...)))? It's implemented in c compilers for ages.

[steven-vd]

it is super easy to forget free(a);.

related to simplifying memory allocation/freeing: X11Libre/xserver#226

being able to use std::unique_ptr or a custom class which would auto-deallocate at the function end would be just indispensable.

It's a dangerous can of worms to open. It can sometimes be convenient, but it also tends to make the code harder to reason about, since you now need the context of the current function plus all of the constructors and destructors that may exist and get silently called.

[powerboat9]

As long as your constructors/destructors don't have weird side effects, that shouldn't be too bad.

[sskras]

@shamefulCake1, thanks for the specifics.

Well, to achieve that instead of C++ I would switch to some newer language maintaining source level compatibility with C. Eg. @ziglang. It can both compile C code and import its' source into own code. But maybe that's another topic.

[steven-vd]

Eg. @ziglang

I like zig, but it's nowhere near mature enough for a project like this

[reaperx7]

C++ requires in code optimizations as opposed to C which uses compile time optimizations. While C++ can produce cleaner code in some instances, the ability to control the compiler for optimizations is more valuable.

[shamefulCake1]

it is super easy to forget free(a);.

related to simplifying memory allocation/freeing: #226

being able to use std::unique_ptr or a custom class which would auto-deallocate at the function end would be just indispensable.

It's a dangerous can of worms to open. It can sometimes be convenient, but it also tends to make the code harder to reason about, since you now need the context of the current function plus all of the constructors and destructors that may exist and get silently called.

I do not disagree a slightest bit, it is a dangerous tool.

But, as has already been mentioned, the cleanup attribute exists in gcc and clang, and is just less famous than unique_ptr, so the problem already exists.

Anyway, I am not qualified to express more of an opinion.

[kkmzero]

The main problem of C++ is that you really need to be authoritarian and have full control over the project to not allow C++ mess. You can write C-like code in C++ and keep it sane but once you have people bothering you with std, virtual functions and templates it's over, the project is finished, it's done. Good luck debugging that mess and have fun spending time rejecting all those donkeys. You might as well propose rewrite to Rust. Limiting standard to C++11 does not help in this case, this is the lesson I had to learn the hard way.

[xgui4]

why ? i didnt know that, i want to more !

[callmetango]

@powerboat9 Thank you for reporting the issue. Due to the broad nature we will move it to the X11Libre · Discussions · GitHub section so that it can be fleshed out further.

[lotuskip]

I meant to never again log into github, but I had to just for this.

[powerboat9]

I meant to never again log into github, but I had to just for this.

[sskras]

Hah, it's a nice satire. :)

[smj-cc]

Reminds me of a similar satire I read somewhere about how K&R wrote Unix and C as a joke.... the punchline was a 200 or so character line of "code" that seriously looked like something a modem would have chucked out with line noise and mismatched baud rate, with the tagline: "We didn't stop until we got this to compile".

[AndersonTorres]

@sjm-cc Unix Haters Handbook?

[smj-cc]

Unix Haters Handbook?

Nope, but thank you for that! I have a copy now.... and it made me go looking. You can find what I was remembering by searching for "Unix is a hoax"

CREATORS ADMIT UNIX, C HOAX

"We stopped when we got a clean compile on the following syntax:
for(;P("\n"),R-;P("|"))for(e=3DC;e-;P("_"+(*u++/8)%2))P("|"+(*u/4)%2); "

Here's one copy: https://www.cs.cmu.edu/~jbruce/humor/unix_hoax.html

[chrisdebian]

Rust?

Edit- Looks like Rust already has a thread

[xgui4]

of course 🤣

[b-aaz]

It's not worth the headache.

[callmetango]

I think the following statement of @metux at [[xlibre] Re: Can a C++ programmer contribute to xlibre? - xlibre - FreeLists] make it very clear: There will be no C++ code in XLibre in the foreseeable future.

[b-aaz]

Nice and concise.