support compiling and linking c++ code

[andrewrk]

Zig supports compiling and linking C code because we already depend on libclang for translate-c. But Clang is also a c++ compiler. It's not really much more work to also support compiling and linking c++ code. I looked into it and it looks like it amounts to about 20k more lines to ship both libc++ and libc++abi.

Note that this does not include translating C++ code.

This would go a long way toward #853, and it would make zig cc even more useful. It also serves the use case of using zig as a build system and package manager for C/C++ projects.

[adontz]

How C++ specific features will work? Like exceptions, for instance.

[jacereda]

I would be quite happy even if it compiled with -fno-exceptions.

[ikskuh]

How C++ specific features will work? Like exceptions, for instance.

I would be quite happy even if it compiled with -fno-exceptions.

It's already possible to build C++ code with exceptions, RTTI and so on with zig today, but you need an external c++ standard library for this. This issue is about providing a c++ standard library with the zig tar ball itself and cross-build that the same way the libc is currently built.

[adontz]

It's already possible to build C++ code with exceptions, RTTI and so on with zig today, but you need an external c++ standard library for this. This issue is about providing a c++ standard library with the zig tar ball itself and cross-build that the same way the libc is currently built.

Yes, but what will be syntax for catching and throwing exceptions? for dynamic_cast?

[ikskuh]

Yes, but what will be syntax for catching and throwing exceptions? for dynamic_cast?

I think your're confusing something here. This issue is not about adding support for interfacing classes and C++ code into Zig (Language), but compiling C++ code with the Zig toolchain. Quote on this:

Note that this does not include translating C++ code.

Which would be required to interface C++ code the same way as C code is translated. And i don't think it should be in scope of the language to ever interface another language than C as it is
humongous work and most languages (including c++) don't really have a stable ABI

[adontz]

What is the point then? C++ code will throw recoverable exception, like ios_base::failure, and crash entire process.

[LakeByTheWoods]

This would (for example) allow a c++ library that exposes a c interface to be built and used in zig.

[adontz]

This would (for example) allow a c++ library that exposes a c interface to be built and used in zig.

Build - yes. I am not sure sure about "used" part. And by "used" I mean not only success path. Zig will have to support C++ exceptions at least to print meaningful stack trace, to say nothing about defer statements as part of stack unwinding.

[ikskuh]

It's not zigs job to fix badly written C++ libraries that violate ABIs by throwing exceptions through C code. Zig does not require stack unwinding for c++ exceptions in the same way as C or any other language does not need this. C++ exceptions are an implementation detail and this should stay like this.

[adontz]

Okay, if that makes sense...