[Bug c++/108965] New: g++: unable to parse c11 _Generics

[Bug c++/108965] New: g++: unable to parse c11 _Generics

[chrisfriedt at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108965

            Bug ID: 108965
           Summary: g++: unable to parse c11 _Generics
           Product: gcc
           Version: 11.3.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c++
          Assignee: unassigned at gcc dot gnu.org
          Reporter: chrisfriedt at gmail dot com
  Target Milestone: ---

G++ version 11.3.0 fails to parse the C11 _Generic keyword. I would suspect
that earlier versions also fail to do this and likely also that later versions
fail as well.

gcc --version
gcc (Ubuntu 11.3.0-1ubuntu1~22.04) 11.3.0
Copyright (C) 2021 Free Software Foundation, Inc.
This is free software; see the source for copying conditions.  There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

It's not clear to me if any part of the ISO C++ standard requires a C++
compiler to parse C11 _Generic, but it certainly can make life less pleasant,
as the equivalent C++ code can sometimes rely on static inline constexpr
templates which are apparently "not cool" until C++14. This particularly
affects people who are using somewhat older C++ toolchains that do not yet
support C++14.

LLVM has no problem, OTOH.

Sample taken from https://en.cppreference.com/w/c/language/generic
```
#include <stdio.h>
#include <math.h>

// Possible implementation of the tgmath.h macro cbrt
#define cbrt(X) _Generic((X), \
              long double: cbrtl, \
                  default: cbrt,  \
                    float: cbrtf  \
              )(X)

int main(void)
{
    double x = 8.0;
    const float y = 3.375;
    printf("cbrt(8.0) = %f\n", cbrt(x)); // selects the default cbrt
    printf("cbrtf(3.375) = %f\n", cbrt(y)); // converts const float to float,
                                            // then selects cbrtf
}
```

```
g++ -o /tmp/main /tmp/main.cpp
/tmp/main.cpp: In function ‘int main()’:
/tmp/main.cpp:6:15: error: expected primary-expression before ‘long’
    6 |               long double: cbrtl, \
      |               ^~~~
/tmp/main.cpp:15:32: note: in expansion of macro ‘cbrt’
   15 |     printf("cbrt(8.0) = %f\n", cbrt(x)); // selects the default cbrt
      |                                ^~~~
/tmp/main.cpp:7:19: error: expected primary-expression before ‘default’
    7 |                   default: cbrt,  \
      |                   ^~~~~~~
/tmp/main.cpp:15:32: note: in expansion of macro ‘cbrt’
   15 |     printf("cbrt(8.0) = %f\n", cbrt(x)); // selects the default cbrt
      |                                ^~~~
/tmp/main.cpp:8:21: error: expected primary-expression before ‘float’
    8 |                     float: cbrtf  \
      |                     ^~~~~
/tmp/main.cpp:15:32: note: in expansion of macro ‘cbrt’
   15 |     printf("cbrt(8.0) = %f\n", cbrt(x)); // selects the default cbrt
      |                                ^~~~
/tmp/main.cpp:5:17: error: ‘_Generic’ was not declared in this scope
    5 | #define cbrt(X) _Generic((X), \
      |                 ^~~~~~~~
/tmp/main.cpp:15:32: note: in expansion of macro ‘cbrt’
   15 |     printf("cbrt(8.0) = %f\n", cbrt(x)); // selects the default cbrt
      |                                ^~~~
/tmp/main.cpp:6:15: error: expected primary-expression before ‘long’
    6 |               long double: cbrtl, \
      |               ^~~~
/tmp/main.cpp:16:35: note: in expansion of macro ‘cbrt’
   16 |     printf("cbrtf(3.375) = %f\n", cbrt(y)); // converts const float to
float,
      |                                   ^~~~
/tmp/main.cpp:7:19: error: expected primary-expression before ‘default’
    7 |                   default: cbrt,  \
      |                   ^~~~~~~
/tmp/main.cpp:16:35: note: in expansion of macro ‘cbrt’
   16 |     printf("cbrtf(3.375) = %f\n", cbrt(y)); // converts const float to
float,
      |                                   ^~~~
/tmp/main.cpp:8:21: error: expected primary-expression before ‘float’
    8 |                     float: cbrtf  \
      |                     ^~~~~
/tmp/main.cpp:16:35: note: in expansion of macro ‘cbrt’
   16 |     printf("cbrtf(3.375) = %f\n", cbrt(y)); // converts const float to
float,
      |                                   ^~~~
```

[Bug c++/108965] g++: unable to parse c11 _Generics

[redi at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108965

--- Comment #1 from Jonathan Wakely <redi at gcc dot gnu.org> ---
(In reply to Christopher Friedt from comment #0)
> It's not clear to me if any part of the ISO C++ standard requires a C++
> compiler to parse C11 _Generic,

It's very clear that it doesn't, _Generic is never mentioned anywhere in the
C++ standard.

> but it certainly can make life less
> pleasant, as the equivalent C++ code can sometimes rely on static inline
> constexpr templates which are apparently "not cool" until C++14. This

Why not? What exactly is the equivalent code you're referring to?

C++ has never needed _Generic because it supports function overloading.

[Bug c++/108965] g++: unable to parse c11 _Generics

[redi at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108965

Jonathan Wakely <redi at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |RESOLVED
         Resolution|---                         |INVALID

--- Comment #2 from Jonathan Wakely <redi at gcc dot gnu.org> ---
I've commented on the github issue. There is no G++ bug here.

[Bug c++/108965] g++: unable to parse c11 _Generics

[chrisfriedt at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108965

--- Comment #3 from Christopher Friedt <chrisfriedt at gmail dot com> ---
All you need to do is look at the example above pulled directly from
cppreference.com, but please simply gaslight the user if that's an easier path
to resolution for you.

[Bug c++/108965] g++: unable to parse c11 _Generics

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108965

--- Comment #4 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
(In reply to Christopher Friedt from comment #3)
> All you need to do is look at the example above pulled directly from
> cppreference.com, but please simply gaslight the user if that's an easier
> path to resolution for you.

There is no gaslighting going on. In fact now you just being abusive for what
end?

The facts are:
The _Generic is only part of the C standard and not part of the C++ standard.
They are two different things.

Gcc's C++ frontend has never implemented _Generic support. GCC's C frontend
does though. 
The C and C++ frontends in gcc do share some code but the parser has never been
shared.

[Bug c++/108965] g++: unable to parse c11 _Generics

[redi at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108965

--- Comment #5 from Jonathan Wakely <redi at gcc dot gnu.org> ---
(In reply to Christopher Friedt from comment #3)
> All you need to do is look at the example above pulled directly from
> cppreference.com, but please simply gaslight the user if that's an easier
> path to resolution for you.

WTF? How am I gaslighting you?

Your example is from cppreference.com/w/c/ which is the **C** reference. Look
at the breadcrumb thingy at the top of the page:

"C / C language / Expressions"


The C++ references are under cppreference.com/w/cpp/ instead of /w/c/

Go to https://en.cppreference.com/w/Main_Page and check that it has both C++
and C references on that site, with different URLs.

[Bug c++/108965] g++: unable to parse c11 _Generics

[chrisfriedt at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108965

--- Comment #6 from Christopher Friedt <chrisfriedt at gmail dot com> ---
It's supported OOTB in `clang++` but fails in `g++`.

The example above is the simplest example that illustrates the issue.

I am not being abusive, but it certainly did feel like gaslighting to read
"you're doing it wrong" / close invalid without any consideration.

Yes, the example is valid C. Yes, the example compiles under `clang++`. No, the
example does not compile under `g++`.

Thanks for following up! Although it's getting kind of condescending now.

[Bug c++/108965] g++: unable to parse c11 _Generics

[redi at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108965

--- Comment #7 from Jonathan Wakely <redi at gcc dot gnu.org> ---
(In reply to Christopher Friedt from comment #6)
> It's supported OOTB in `clang++` but fails in `g++`.

Nobody is disputing that, but Clang supports lots of things that aren't valid
in C++ and aren't supported by G++, and G++ supports things that aren't valid
in C++ and aren't supported by Clang. That's not a bug.

> The example above is the simplest example that illustrates the issue.

Yes, and it's not valid C++.

> I am not being abusive, but it certainly did feel like gaslighting to read
> "you're doing it wrong" / close invalid without any consideration.

I commented explaining that it's not in the C++ standard, and followed up in
the github issue where the real bug is. That's hardly "without consideration".
I've shown how to fix your C++ code to make it valid in C++11, and pointed out
a problem in your C code using _Generic. I didn't have to do any of that, and I
certainly won't bother doing so again.

[Bug c++/108965] g++: unable to parse c11 _Generics

[chrisfriedt at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108965

--- Comment #8 from Christopher Friedt <chrisfriedt at gmail dot com> ---
My code is clearly valid C++ according to g++ :-)

Thanks for your help in any case.

On Tue, Feb 28, 2023, 7:38 AM redi at gcc dot gnu.org <
gcc-bugzilla@gcc.gnu.org> wrote:

> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108965
>
> --- Comment #7 from Jonathan Wakely <redi at gcc dot gnu.org> ---
> (In reply to Christopher Friedt from comment #6)
> > It's supported OOTB in `clang++` but fails in `g++`.
>
> Nobody is disputing that, but Clang supports lots of things that aren't
> valid
> in C++ and aren't supported by G++, and G++ supports things that aren't
> valid
> in C++ and aren't supported by Clang. That's not a bug.
>
> > The example above is the simplest example that illustrates the issue.
>
> Yes, and it's not valid C++.
>
> > I am not being abusive, but it certainly did feel like gaslighting to
> read
> > "you're doing it wrong" / close invalid without any consideration.
>
> I commented explaining that it's not in the C++ standard, and followed up
> in
> the github issue where the real bug is. That's hardly "without
> consideration".
> I've shown how to fix your C++ code to make it valid in C++11, and pointed
> out
> a problem in your C code using _Generic. I didn't have to do any of that,
> and I
> certainly won't bother doing so again.
>
> --
> You are receiving this mail because:
> You reported the bug.

[Bug c++/108965] g++: unable to parse c11 _Generics

[redi at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108965

--- Comment #9 from Jonathan Wakely <redi at gcc dot gnu.org> ---
(In reply to Christopher Friedt from comment #8)
> My code is clearly valid C++ according to g++ :-)

Maybe you mean clang++ but even then, no it's not:


$ clang++ -pedantic gen.cc
gen.cc:15:32: warning: '_Generic' is a C11 extension [-Wc11-extensions]
    printf("cbrt(8.0) = %f\n", cbrt(x)); // selects the default cbrt
                               ^
gen.cc:5:17: note: expanded from macro 'cbrt'
#define cbrt(X) _Generic((X), \
                ^
gen.cc:16:35: warning: '_Generic' is a C11 extension [-Wc11-extensions]
    printf("cbrtf(3.375) = %f\n", cbrt(y)); // converts const float to float,
                                  ^
gen.cc:5:17: note: expanded from macro 'cbrt'
#define cbrt(X) _Generic((X), \
                ^
2 warnings generated.


It's a non-standard extension, not valid C++.

[Bug c++/108965] g++: unable to parse c11 _Generics

[chrisfriedt at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108965

--- Comment #10 from Christopher Friedt <chrisfriedt at gmail dot com> ---
Thanks - I wasn't using -pedantic, but you have certainly proven stuff.

[Bug c++/108965] g++: unable to parse c11 _Generics

[chrisfriedt at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108965

--- Comment #11 from Christopher Friedt <chrisfriedt at gmail dot com> ---
(In reply to Jonathan Wakely from comment #9)
> (In reply to Christopher Friedt from comment #8)
> > My code is clearly valid C++ according to g++ :-)
>
> Maybe you mean clang++ but even then, no it's not:

I was talking about the code at the linked PR. It's valid C++11 according to
g++ (and clang++).

I was not talking about the cppreference.com example (which is not my code,
which is what you seem to be referencing as my code).

I did not compile the cppreference.com code with -pedantic, but if you feel you
need to in order to illustrate that you are right here, by all means.

I can also run a command that illustrates my point as well. See?

$ clang++ -std=c++11 -o /tmp/main /tmp/main.cpp
$ echo $?
0

So really, if you're trying to tell me that my observations did not actually
occur, that is, by definition, gaslighting.

> I commented explaining that it's not in the C++ standard, and followed up
> in the github issue where the real bug is.

The comments you made on the github pr do not highlight "where the real bug is"
because the code is not buggy - it produces the desired results. 

> "I didn't have to do any of that, and I certainly won't bother doing so again"

Works for me!