On: C++ in the Linux Kernel

[monocasa]

Lol, part of me likes the effort taken just because, but the kernel devs _really_ do not want C++. One hint: "struct class"

https://elixir.bootlin.com/linux/latest/source/include/linux...

[nly → monocasa]

What bothers me about that is that, because C doesn't have namespaces, it's already a terrible name for a struct. What if you want another "class" of thing?

Call it device_class ffs

[tuyiown → nly]

You're dismissing the fact that the keyword collision really well might be intentional, the worst of it is that `/sys/class` siblings `bus` and `driver`, if their internal linux rep is actually in the `class.h` siblings, are called `struct bus_type` and `struct device_driver`

[leeter → tuyiown]

I'd say it was extremely intentional given this:

https://lwn.net/ml/linux-api/20180905165436.GA25206@kroah.co...

And that's for a userspace header.

[steerablesafe → monocasa]

   #define class Class
   #include <some_linux_header.h>
   #undef class

/s

[Pinus → monocasa]

Wasn't there a period recently — around 1995 or so — when the Linux kernel had to be compiled with a C++ compiler?

[cesarb → Pinus]

Quoting from http://vger.kernel.org/lkml/#s15-3 :

"In the dark old days, in the time that most of you hadn't even heard of the word "Linux", the kernel was once modified to be compiled under g++. That lasted for a few revisions. People complained about the performance drop. It turned out that compiling a piece of C code with g++ would give you worse code. It shouldn't have made a difference, but it did. Been there, done that."

[bluGill → cesarb]

I wonder if it is still true. C++ compilers have come a long way. (as have C compilers). C++ is 99% a superset of C, I'm not sure how much of that last is used in the kernel, so it might be too much effort, but C++ is in a few cases stricter than C in ways that compilers can use to optimize. Many C programs run faster when compiled in C++ these days.

If there is a difference (either way) I'd expect it to be something you can measure, but not something you would notice in the real world on one computer. (though at google scale it probably shows up)

[monocasa → Pinus]

I don't think that's accurate. Maybe you're thinking of how it needs gcc extensions?

[mijoharas → Pinus]

Why was that the case?

[jerrysievert → monocasa]

as long as nothing you're including includes that in c++, it shouldn't be an issue at the linker level, and thus not be an issue at all.

at least in general - if it's something that can't handled by a c shim then you might have an issue.

[monocasa → jerrysievert]

It's one of the most fundamental structures in the kernel. Pretty much all driver headers include it, if indirectly.