Home | Manual | Playground

Cake - C23 and Beyond

Intro

Cake works as an extension for MSVC on Windows and as an extension for GCC on Linux. This approach makes Cake useful in real and existing programs.

When applicable, Cake uses the same command line options of MSVC and GCC.

Static analyzer

For static analyzer concepts of ownership and nullable pointers visit ownership

Include directories

On Windows, Cake can be used on the command line similarly to MSVC. Cake reads the INCLUDE variable, the same variable used by MSVC to locate the include directories.

Additionally, you can run Cake outside the Visual Studio command prompt by placing the file cakeconfig.h in the same directory or above the source files, and specifying the directories using #pragma dir.

If Cake doesn't find cakeconfig.h in the local directories, it will try to locate it in the same path as the Cake executable.

The -autoconfig option generates the cakeconfig.h automatically on both Windows and Linux.

To manually discover which directories are included, you can run the command:

echo %INCLUDE%

at Visual Studio command prompt.

To find out what are the directories used by GCC type:

echo | gcc -E -Wp,-v -

Sample of cakeconfig.h


#ifdef __linux__
/*
   To find the include directories used my GCC type:   
   echo | gcc -E -Wp,-v -
*/
#pragma dir "/usr/lib/gcc/x86_64-linux-gnu/11/include"
#pragma dir "/usr/local/include"
#pragma dir "/usr/include/x86_64-linux-gnu"
#pragma dir "/usr/include"

#endif

#ifdef _WIN32
/*
   To find the include directories used my  MSVC,
   open Visual Studio Developer Commmand prompt and type:
   echo %INCLUDE%.
   Running Cake inside mscv command prompt uses %INCLUDE% automatically.
*/
#pragma dir "C:/Program Files/Microsoft Visual Studio/2022/Professional/VC/Tools/MSVC/14.38.33130/include"
#pragma dir "C:/Program Files/Microsoft Visual Studio/2022/Professional/VC/Tools/MSVC/14.38.33130/ATLMFC/include"
#pragma dir "C:/Program Files/Microsoft Visual Studio/2022/Professional/VC/Auxiliary/VS/include"
#pragma dir "C:/Program Files (x86)/Windows Kits/10/include/10.0.22000.0/ucrt"
#pragma dir "C:/Program Files (x86)/Windows Kits/10/include/10.0.22000.0/um"
#pragma dir "C:/Program Files (x86)/Windows Kits/10/include/10.0.22000.0/shared"
#pragma dir "C:/Program Files (x86)/Windows Kits/10/include/10.0.22000.0/winrt"
#pragma dir "C:/Program Files (x86)/Windows Kits/10/include/10.0.22000.0/cppwinrt"
#pragma dir "C:/Program Files (x86)/Windows Kits/NETFXSDK/4.8/include/um"

#endif

Sample, project cakeconfig.h


//system includes...etc
#include "C:\Program Files (x86)\cake\cakeconfig.h"

//project extra includes
#pragma dir ".\openssl\include"

Command line

cake [options] source1.c source2.c ...

SAMPLES

    cake source.c
    Compiles source.c and outputs /out/source.c

    cake -target=C11 source.c
    Compiles source.c and outputs C11 code at /out/source.c

cake file.c -o file.cc && cl file.cc
    Compiles file.c and outputs file.cc then use cl to compile file.cc

cake file.c -direct-compilation -o file.cc && cl file.cc
    Compiles file.c and outputs file.cc for direct compilation then use cl to compile file.cc

OPTIONS

-I (same as GCC and MSVC)

Adds a directory to the list of directories searched for include files

-no-output

Cake will not generate output

-D (same as GCC and MSVC)

Defines a preprocessing symbol for a source file

-E (same as GCC and MSVC)

Copies preprocessor output to standard output

-o name.c (same as GCC and MSVC)

Defines the output name. used when we compile one file

-remove-comments

Remove all comments from the output file

-direct-compilation

output code as compiler sees it without macros.

-target=standard

Output target C standard (c89, c99, c11, c23, c2y, cxx) C99 is the default and C89 (ANSI C) is the minimum target

-dump-tokens

Output tokens before preprocessor

-fi

Format input (format before language conversion)

-fo

Format output (format after language conversion, result parsed again)

-ir

Generates C89 code

-Wname -Wno-name (same as GCC)

Enables or disable warnings. See warnings

-disable-assert

disable cake extension where assert is an statement. See extensions

-showIncludes

Causes the compiler to output a list of the include files. The option also displays nested include files, that is, the files included by the files that you include.

-Wall

Enables all warnings

-sarif

Generates sarif files. Sarif Visual Studio plugin https://marketplace.visualstudio.com/items?itemName=WDGIS.MicrosoftSarifViewer

-sarif-path

Specifies the Sarif output dir.

Inside "Visual Studio -> External Tools" this command can be used for static analysis.

`-Wstyle -msvc-output -no-output -sarif -sarif-path "$(SolutionDir).sarif" $(ItemPath)´

-msvc-output

Output is compatible with visual studio IDE. We can click on the error message and IDE selects the line.

-fanalyzer

This option enables an static analysis of program flow. This is required for some ownership checks

-auto-config

Generates cakeconfig.h header.

On Windows, it must be generated inside the Visual Studio Command Prompt to read the INCLUDE variable. On Linux, it calls GCC with echo | gcc -v -E - 2>&1 and reads the output.

Output

One directory called out is created keeping the same directory structure of the input files.

For instance:

cake c:\project\file1.c

output:

  c:\project
  ├── file1.c
  ├── out
      ├── file1.c

More files..

cake c:\project\file1.c c:\project\other\file2.c

output

  c:\project
  ├── file1.c
  ├── other
  │   ├── file2.c
  ├── out
      ├── file1.c
      ├── other
          ├── file2.c

Pre-defined macros

 #define __CAKE__ 202311L
 #define __STDC_VERSION__ 202311L
 #define __STDC_OWNERSHIP__ 1

The define STDC_OWNERSHIP indicates that the compiler suports owneship checks

Pre-defined macros for MSVC compatibility

https://learn.microsoft.com/en-us/cpp/preprocessor/predefined-macros?view=msvc-170#standard-predefined-macros

Pre-defined macros for GCC compatibility

https://gcc.gnu.org/onlinedocs/cpp/Predefined-Macros.html

C99 Transformations

C89 https://port70.net/~nsz/c/c89/c89-draft.html

C99 https://open-std.org/JTC1/SC22/WG14/www/docs/n1124.pdf

 #define __STDC_VERSION__ 199901L  //C99

C99 restrict pointers

void f(const char* restrict s);

Becomes in C89

void f(const char* /*restrict*/ s);

N448

C99 Variable-length array (VLA)

The idea is not implement variable length arrays with automatic storage duration. (__STDC_NO_VLA__ 1).

But there are other uses of VLA.

#include <stdlib.h>
#include <stdio.h>

int main() {
    int n = 2;
    int m = 3;
    int (*p)[n][m] = malloc(sizeof * p);

    printf("%zu\n", sizeof(*p));

    free(p);
}

Becomes C89 (not implemented)

#include <stdlib.h>
#include <stdio.h>

int main() {
    int n = 2;
    int m = 3;
    
    /*these variables are created to store the dynamic size*/
    const int vla_1_n = n;
    const int vla_1_m = m;
    
    int (*p)[n][m] = malloc((vla_1_n*vla_1_m)*sizeof(int));

    printf("%zu\n", (vla_1_n*vla_1_m)*sizeof(int));

    free(p);
}

https://www.open-std.org/jtc1/sc22/wg14/www/docs/n683.htm

C99 Flexible array members

struct s {
    int n;
    double d[]; 
};

Becomes (not implemented)

struct s {
    int n;
    double d[1];
};

//TODO sizeof

C99 static and type qualifiers in parameter array declarators

#include <stdlib.h>

void F(int a[static 5]) {
}

int main() 
{    
    F(0);
    F(NULL);
    F(nullptr);

    int a[] = {1, 2, 3};    
    F(a);//error
    
    int b[] = { 1, 2, 3 , 4, 5};
    F(b); 

    int c[] = { 1, 2, 3 , 4, 5, 6};
    F(c);
}

Cake verifies that the argument is an array with at least the specified number of elements. It extends this check to arrays without a static size as well.

C99 Complex and imaginary support

Not implemented

C99 Universal character names (\u and \U)

Not implemented

C99 Hexadecimal floating constants

double d = 0x1p+1;

Becomes in C89

double d = 2.000000;

Cake converts hexadecimal floating-point values to decimal floating-point representation using strtod followed by snprintf. This conversion may introduce precision loss.

C99 Compound literals

struct s {
  int i;
};

int f(void) {
  struct s * p = 0, * q;
  int j = 0;
  again:
    q = p, p = & ((struct s) { j++ });
  if (j < 2) goto again;
  return p == q && q -> i == 1;
}

Becomes in C89

struct s {
  int i;
};
int f(void) {
  struct s * p = 0, * q;
  int j = 0;
  again:
    struct s compound_literal_1 = { j++ };
    q = p, p = & compound_literal_1;
  if (j < 2) goto again;
  return p == q && q -> i == 1;
}

N716 https://www.open-std.org/jtc1/sc22/wg14/www/docs/n716.htm

C99 Designated initializers

 int main()
 {
  int a[6] = {[4] = 29, [2] = 15 };

  struct point { int x, y; };

  struct point p = { .y = 2, .x = 3 }
 }

Becomes C89

int main()
{
  int a[6] = { 0, 0, 15, 0, 29, 0 };
  struct point { int x, y; };
  struct point p = { 3, 2 }
}

N494 https://www.open-std.org/jtc1/sc22/wg14/www/docs/n494.pdf

C99 Line comments

When compiling to C89 line comments are converted to /comments/.

C99 inline functions

TODO https://www.open-std.org/jtc1/sc22/wg14/www/docs/n741.htm

C99 _Pragma preprocessing operator

TODO

C99 __func__ predefined identifier

Parsed. C89 conversion not implemented yet.

C99 Variadic macros

We need to expand the macro when comping to C89. This is covered by # macro expand.

Sample:


#include <stdio.h>

#define debug(...) fprintf(stderr, __VA_ARGS__)
#pragma expand debug

int main()
{
  int x = 1;
  debug("X = %d\n", 1);
}

Becomes

#include <stdio.h>

#define debug(...) fprintf(stderr, __VA_ARGS__)
#pragma expand debug

int main()
{
  int x = 1;
  fprintf(stderr, "X = %d\n", 1);
}

I am considering to mark the debug macro to be expanded automatically if __VA_ARGS__ is used. Then pragma expand will not be necessary.

N707 https://www.open-std.org/jtc1/sc22/wg14/www/docs/n707.htm

C99 _Bool

When compiling to C89 _Bool is replaced by unsigned char.

//line comments
int main(void)
{
    _Bool b = 1;
    return 0;
}

Becomes in C89

/*line comments*/
int main(void)
{
    unsigned char b = 123;
    return 0;
}

Alternative design - typedef ? Considering C23 has bool and the objective of C89 version is to have a version that compiles in C++ the best option would be use bool, true, false.

Obs: Currently cake is not converting 123 to 1 as required by C standard.

C11 Transformations

#define __STDC_VERSION__ 201112L //C11

https://open-std.org/JTC1/SC22/WG14/www/docs/n1570.pdf

https://files.lhmouse.com/standards/ISO%20C%20N2176.pdf

C11 _Static_assert

When compiling to versions < C11 _Static_Assert is removed.

C11 Anonymous structures and unions

It is implemented, however the conversion to C99, C89 was not implemented yet.

struct v {
  union { /* anonymous union*/
     struct { int i, j; }; /* anonymous structure*/
     struct { long k, l; } w;
  };
  int m;
} v1;

int main(){
  v1.i = 2; /* valid*/
  v1.w.k = 5; /* valid*/
}

C11 _Noreturn

_Noreturn void f () {
  abort(); // ok
}

Becomes in < C11

/*[[noreturn]]*/ void f () {
  abort(); // ok
}

C23 attribute [[noreturn]] provides similar semantics. The _Noreturn function specifier is an obsolescent feature

C11 Thread_local/Atomic

Parsed but not transformed.

C11 type-generic expressions (_Generic)

When compiling to C99, C89 we keep the expression that matches the type.

For instance:

The expression that matches the argument 1.0 is cbrtl.

The result of _Generic in C99 will be cbrtl. Because this is inside a macro we need to tell the transpiler to expand that macro using pragma expand.

N1441 https://www.open-std.org/jtc1/sc22/wg14/www/docs/n1441.htm

#include <math.h>

#define cbrt(X) _Generic((X),    \
                  double: cbrtl, \
                  float: cbrtf , \
                  default: cbrtl \
              )(X)

#pragma expand cbrt

int main(void)
{
    cbrt(1.0);
}

Becomes in C99, C89

#include <math.h>

#define cbrt(X) _Generic((X),    \
                  double: cbrtl, \
                  float: cbrtf , \
                  default: cbrtl \
              )(X)

#pragma expand cbrt

int main(void)
{
     cbrtl(1.0);
}

C11 u' ' U' ' character constants

 int i = U'ç';
 int i2 = u'ç';

Becomes in < C11

 int i = 231u;
 int i2 = ((unsigned short)231);

Important: Cake assume source is utf8 encoded.

C11 u8"literals"

u8 literals are converted to escape sequences.

char * s1 = u8"maçã";
char * s2 = u8"maca";

Becomes in < C11

char * s1 = "ma\xc3\xa7\xc3\xa3";
char * s2 = "maca";

N1488 https://www.open-std.org/jtc1/sc22/wg14/www/docs/n1488.htm

Important: Cake assume source is utf8 encoded.

C11 _Alignof or C23 alignof

When compiling to C99 or C89 it is replaced by the equivalent constant.

 int main()
 {
   int align = alignof(int);
 }

Becomes < C11

 int main()
 {
   int align = 4;
 }

TODO considering a macro. For instance, ALIGNOF_INT

C11 _Alignas or C23 alignas

Not implemented.

C23 Transformations

https://open-std.org/JTC1/SC22/WG14/www/docs/n3096.pdf

#define __STDC_VERSION__ 201710L  //C17
#define __STDC_VERSION__ 202311L  //C23

C23 _Decimal32, _Decimal64, and _Decimal128

Not implemented. https://www.open-std.org/jtc1/sc22/wg14/www/docs/n1107.htm

C23 static_assert / single-argument static_assert

In C23 static_assert is a keyword and the text message is optional.

Whe comping to C11, static_assert is replaced by it C11 version _Static_assert. If the static_assert has only one argument the text becomes "error".

N1330 https://www.open-std.org/jtc1/sc22/wg14/www/docs/n1330.pdf

int main()
{    
    static_assert(1 == 1, "error message");
    static_assert(1 == 1);
}

Becomes in C11

int main()
{    
    _Static_assert(1 == 1, "error message");
    _Static_assert(1 == 1, "error");
}

In < C11 it is replaced by one space;

C23 u8 character prefix

Implemented. https://open-std.org/JTC1/SC22/WG14/www/docs/n2418.pdf

int main(){
    unsigned char c = u8'~';
}

When compiling to < C23 becomes.

int main(){
    unsigned char c = ((unsigned char)'~');
}

C23 No function declarators without prototypes

Implemented. https://www.open-std.org/JTC1/SC22/WG14/www/docs/n2841.htm

int main(){
    func(); //this is an error in C23
}

See also Remove support for function definitions with identifier lists

https://open-std.org/JTC1/SC22/WG14/www/docs/n2432.pdf

C23 Improved Tag Compatibility

Not implemented yet.

https://www.open-std.org/jtc1/sc22/wg14/www/docs/n3037.pdf

struct foo { int a; } p;
void bar(void)
{
  struct foo { int a; } q;
  q = p;
}

Becomes < C23

struct foo { int a; } p;
void bar(void)
{
  struct foo  q;
  q = p;
}

C23 Unnamed parameters in function definitions

int f(int );

int f(int ) {
}

https://open-std.org/JTC1/SC22/WG14/www/docs/n2480.pdf

Cake should add a dummy name when generating C < 23. (Not implemented yet)

C23 Digit separators

int main()
{
    int a = 1000'00;
}

Becomes in < C23

int main()
{
    int a = 100000;
}

This transformation happens at token level, so even preprocessor and inactive blocks are transformed.

https://www.open-std.org/jtc1/sc22/wg14/www/docs/n2626.pdf

C23 Binary literals

#define X  0b1010

int main()
{
    int a = X;
    int b = 0B1010;
}

Becomes in C11, C99, C89

#define X  0xa

int main()
{
    int a = X;
    int b = 0xa;
}

This transformation happens at token level, so even preprocessor and inactive blocks are transformed.

C23 Introduce the nullptr constant


int main()
{
  void * p = nullptr;
  auto p2 = nullptr;
  typeof(nullptr) p3 = nullptr;
}

Becomes in < C23

int main()
{
  void * p = ((void*)0);
  void  * p2 = ((void*)0);
  void  * p3 = ((void*)0);
}

https://open-std.org/JTC1/SC22/WG14/www/docs/n3042.htm

C23 Make false and true first-class language features

When compiling to C89 bool is replaced by unsigned char, true by 1 and false by 0.

When compiling to C99 and C11 bool is replaced with _Bool, true is replaced with ((_Bool)1) and false with (_Bool)0)

https://www.open-std.org/jtc1/sc22/wg14/www/docs/n2935.pdf

C23 {} empty initializer


int main()
{
    struct X {
        int i;
    } x = {};

    x = (struct X) {};

    struct Y
    {
        struct X x;
    } y = { {} };
}

Becomes in < C23


int main()
{
    struct X {
        int i;
    } x = {0};

    x = (struct X) {0};

    struct Y
    {
        struct X x;
    } y = { {0} };
}

Note: Cake code is 100% equivalent because it does not make padding bit zero.

C23 auto

static auto a = 3.5;
auto p = &a;

double A[3] = { 0 };
auto pA = A;
auto qA = &A;

Becomes < C23

static double a = 3.5;
double  * p = &a;

double A[3] = { 0 };
double  * pA = A;
double  (* qA)[3] = &A;

https://open-std.org/JTC1/SC22/WG14/www/docs/n3007.htm

C23 typeof / typeof_unqual


#define SWAP(a, b) \
  do {\
    typeof(a) temp = a; a = b; b = temp; \
  } while (0)

#pragma expand SWAP

int main()
{
    /*simple case*/
    int a = 1;
    typeof(a) b = 1;

    /*pay attention to the pointer*/
    typeof(int*) p1, p2;

    /*let's expand this macro and see inside*/
    SWAP(a, b);

    /*for anonymous structs we insert a tag*/
    struct { int i; } x;
    typeof(x) x2;
    typeof(x) x3;

   /*Things get a little more complicated*/
   int *array[2];
   typeof(array) a1, a2;
   
   typeof(array) a3[3];
   typeof(array) *a4[4];

   /*abstract declarator*/
   int k = sizeof(typeof(array));

   /*new way to declare pointer to functions?*/
   typeof(void (int)) * pf = NULL;
}

Becomes in < C23



#define SWAP(a, b) \
  do {\
    typeof(a) temp = a; a = b; b = temp; \
  } while (0)

#pragma expand SWAP

int main()
{
    /*simple case*/
    int a = 1;
    int  b = 1;

    /*pay attention to the pointer*/
    int  *p1,  *p2;

    /*let's expand this macro and see inside*/
     do {int temp = a; a = b; b = temp; } while (0);

    /*for anonymous structs we insert a tag*/
    struct _anonymous_struct_0 { int i; } x;
    struct _anonymous_struct_0  x2;
    struct _anonymous_struct_0  x3;

   /*Things get a little more complicated*/
   int *array[2];
   int  *a1[2],  *a2[2];
   
   int  *(a3[3])[2];
   int  *(*a4[4])[2];

   /*abstract declarator*/
   int k = sizeof(int*[2]);
   
   /*new way to declare pointer to functions?*/
   void  (*pf)(int) = ((void*)0);
}

https://open-std.org/JTC1/SC22/WG14/www/docs/n2927.htm https://open-std.org/JTC1/SC22/WG14/www/docs/n2930.pdf

C23 Improved Normal Enumerations

//TODO

https://open-std.org/JTC1/SC22/WG14/www/docs/n3029.htm

enum a {
    a0 = 0xFFFFFFFFFFFFFFFFULL
};

static_assert(_Generic(a0,
        unsigned long long: 0,
        int: 1,
        default: 2 == 0));

The type of the enum must be adjusted.

C23 constexpr

https://www.open-std.org/jtc1/sc22/wg14/www/docs/n3018.htm

Cake convert constexpr declarator with a cast and the value. addressof constexpr declarator is not implemented.

#include <stdio.h>

constexpr int c = 123;

int a[c];

constexpr double PI = 3.14;

static_assert(PI + 1 == 3.14 + 1.0);

int main()
{
   printf("%f", PI);
}

Becomes < C23

#include <stdio.h>

const int c = 123;

int a[((int)123)];

const double PI = 3.14;

int main()
{
   printf("%f", ((double)3.140000));
}

TODO: Maybe suffix like ULL etc makes the code easier to read.

C23 Enhancements to Enumerations

enum X : short {
  A
};

int main() {
   enum X x = A;   
}

Becomes < C23

enum X {
  A
};

int main() {
   short x = ((short)A);   
}

https://www.open-std.org/jtc1/sc22/wg14/www/docs/n3030.htm

C23 Attributes

Conversion to < C23 will just remove the attributes.

https://www.open-std.org/jtc1/sc22/wg14/www/docs/n2335.pdf https://www.open-std.org/jtc1/sc22/wg14/www/docs/n2554.pdf

Related: Standard Attributes in C and C++ - Timur Doumler - ACCU 2023 https://youtu.be/EpAEFjbTh3I

C23 fallthrough attribute

Not implemented

https://open-std.org/JTC1/SC22/WG14/www/docs/n2408.pdf

C23 deprecated attribute

Partially implemented https://www.open-std.org/jtc1/sc22/wg14/www/docs/n2334.pdf

C23 maybe_unused attribute

Implemented https://www.open-std.org/jtc1/sc22/wg14/www/docs/n2270.pdf

C23 nodiscard attribute

Partially implemented

https://www.open-std.org/jtc1/sc22/wg14/www/docs/n2267.pdf

https://open-std.org/JTC1/SC22/WG14/www/docs/n2448.pdf

C23 [[unsequenced]] and [[reproducible]]

//TODO

https://open-std.org/JTC1/SC22/WG14/www/docs/n2956.htm

C23 __has_attribute

Its is implemented in cake. Conversion < C23 not defined. Maybe a define.

C23 __has_include


#if __has_include(<stdio.h>)
#warning  YES
#endif

#if __has_include(<any.h>)
#warning  YES
#else
#warning  NO
#endif

Its is implemented in cake. Conversion < C23 not defined.

C23 #warning

When compiling to versions < 23 the line is commented out.

int main()
{
  #warning my warning message  
}

When target < C23 becomes

int main()
{
  /* #warning my warning message */  
}

https://www.open-std.org/jtc1/sc22/wg14/www/docs/n2686.pdf

C23 #embed

Partially implemented.

#include <stdio.h>

int main()
{
  static const char file_txt[] = {
   #embed "stdio.h"
   ,0
  };

  printf("%s\n", file_txt);
}

Becomes in < C23


https://www.open-std.org/jtc1/sc22/wg14/www/docs/n3017.htm

#include <stdio.h>

int main()
{
  static const char file_txt[] = {
    35,112,114,/*lot more here ...*/ 13,10
   ,0
  };

  printf("%s\n", file_txt);
}

I am considering add an option to generate a file with a suffix like "embed_stdio.h" then the equivalent code will be:

Becomes in < C23

#include <stdio.h>

int main()
{
  static const char file_txt[] = {
   #include "embed_stdio.h"
   ,0
  };

  printf("%s\n", file_txt);
}

C23 #elifdef #elifndef

#define Y

#ifdef X
#define VERSION 1
#elifdef  Y
#define VERSION 2
#else
#define VERSION 3
#endif

Becomes < C23

#define Y

#ifdef X
#define VERSION 1
#elif defined   Y
#define VERSION 2
#else
#define VERSION 3
#endif

C23 __VA_OPT__

Implemented. Requires #pragma expand.


#define F(...) f(0 __VA_OPT__(,) __VA_ARGS__)
#define G(X, ...) f(0, X __VA_OPT__(,) __VA_ARGS__)
#define SDEF(sname, ...) S sname __VA_OPT__(= { __VA_ARGS__ })
#define EMP

/*maybe this could be automatic if <C23*/
#pragma expand F
#pragma expand G
#pragma expand SDEF
#pragma expand EMP

void f(int i, ...) {}


int main()
{
  int a = 1;
  int b = 2;
  int c = 3;
  
  F(a, b, c);
  F();
  F(EMP);
  G(a, b, c);
  G(a, );
  G(a);

}

Becomes in < C23


#define F(...) f(0 __VA_OPT__(,) __VA_ARGS__)
#define G(X, ...) f(0, X __VA_OPT__(,) __VA_ARGS__)
#define SDEF(sname, ...) S sname __VA_OPT__(= { __VA_ARGS__ })
#define EMP

/*maybe this could be automatic if <C23*/
#pragma expand F
#pragma expand G
#pragma expand SDEF
#pragma expand EMP

void f(int i, ...) {}


int main()
{
  int a = 1;
  int b = 2;
  int c = 3;
  
   f(0, a, b, c);
   f(0 );
   f(0);
   f(0, a, b, c);
   f(0, a );
   f(0, a );

}

https://www.open-std.org/jtc1/sc22/wg14/www/docs/n3033.htm

C23 BitInt(N))

Not implemented

C23 Compound Literals with storage specifier

Not implemented yet.

void F(int *p){}

int main()
{
   F((static int []){1, 2, 3, 0})
}

Becomes (not implemented yet)

void F(int *p){}

int main()
{
    static int _compound_1[] = {1, 2, 3, 0};
    F(_compound_1);
x   }

https://www.open-std.org/jtc1/sc22/wg14/www/docs/n3038.htm

C23 Variably-modified (VM) types

https://www.open-std.org/jtc1/sc22/wg14/www/docs/n2778.pdf

C2Y Transformations

Obsolete implicitly octal literals


static_assert(0o52 == 052);
static_assert(0O52 == 052);
static_assert(0O52 == 42);

int main()
{
    int i = 0o52;
}

Becomes in < C2Y (prefix is removed)


static_assert(052 == 052);
static_assert(052 == 052);
static_assert(052 == 42);

int main()
{
    int i = 052;
}

Extension - defer

https://www.open-std.org/jtc1/sc22/wg14/www/docs/n3199.htm

defer will call the defer statement before the block exit at inverse order of declaration.

     defer-statement:
        defer secondary-block

For instance:

#include <stdio.h>

int main() {
  do {
     FILE* f = fopen("in.txt", "r");
     if (f == NULL) break;
     defer fclose(f);

     FILE* f2 = fopen("out.txt", "w");
     if (f2 == NULL) break;
     defer fclose(f2);
     //...    
  }
  while(0);
}

Becomes in < C2Y

#include <stdio.h>

int main() {
  do {
     FILE* f = fopen("in.txt", "r");
     if (f == ((void*)0)) break;

     FILE* f2 = fopen("out.txt", "w");
     if (f2 == ((void*)0)) {  fclose(f); break;}
     
     fclose(f2); fclose(f);
   }
  while(0);
}

Extension - if with initializer

https://www.open-std.org/jtc1/sc22/wg14/www/docs/n3196.htm

#include <stdio.h>
int main()
{
   int size = 10;
   if (FILE* f = fopen("file.txt", "r"); f)
   {
     /*...*/
     fclose(f);
   }
}

Becomes in < C2Y

#include <stdio.h>

int main()
{
   int size = 10;
   {FILE* f = fopen("file.txt", "r");if ( f)
   {
     /*...*/
     fclose(f);
   }}
}

C++ proposal https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2016/p0305r0.html

Extension typename on _Generic

This feature was created in Cake and now it is part of C2Y!

https://www.open-std.org/jtc1/sc22/wg14/www/docs/n3260.pdf

 int main()
{
    const int * const p;
    static_assert(_Generic(p, const int *: 1));

    /*extension*/
    static_assert(_Generic(int, int : 1));
    static_assert(_Generic(typeof(p), const int * const: 1));
}

Cake Extensions (Not in C23, C2Y)

Extension - try catch throw

   try-statement:
      try secondary-block
      try secondary-block catch secondary-block   

jump-statement:
  throw;

try catch is a external block that we can jump off.

try catch is a LOCAL jump this is on purpose not a limitation.

catch block is optional.

try
{
   for (int i = 0 ; i < 10; i++) {
      for (int j = 0 ; j < 10; j++) {
        ... 
        if (error) throw;
        ...
      }
   }
}
catch
{
}

Extension Literal function - lambdas

Lambdas without capture where implemented using a syntax similar of compound literal for function pointer.

Lambdas are the most complex code transformation so far because sometimes function scope types needs to be transformed to file scope. This is important because manual lambda capture is something we want to use in function scope.

For instance:

extern char* strdup(const char* s);
void create_app(const char* appname)
{
  struct capture {
     char * name;
  } capture = { .name = strdup(appname) };

  (void (void* p)) {
    struct capture* capture = p;    
  }(&capture); 
}

Because struct capture was in function scope and the lambda function will be created at file scope the type struct capture had to be moved from function scope to file scope.

extern char* strdup(const char* s);

struct _capture0 {
     char * name;
  };
  
void _lit_func_0(void *p) {
    struct _capture0* capture = p;    
  }

void create_app(const char* appname)
{
  struct _capture0  capture = { .name = strdup(appname) };
  _lit_func_0(&capture);  
}

Extension #pragma dir

#pragma dir "C:/Program Files (x86)/Windows Kits/10//include/10.0.22000.0/cppwinrt"

pragma dir makes the preprocessor include the directory when searching for includes.

Extension #pragma expand

pragma expand tells the C back-end to not hide macro expansions. This is necessary when the compiler makes changes inside macro expanded code.

For instance:


#define SWAP(a, b) \
    do { \
      typeof(a) temp = a; a = b; b = temp; \
    } while(0)

#pragma expand SWAP

int main()
{
   int a = 1;
   typeof(a) b = 2;
   SWAP(a, b);
   return 1;
}

Becomes

#define SWAP(a, b) \
    do { \
      typeof(a) temp = a; a = b; b = temp; \
    } while(0)

#pragma expand SWAP

int main()
{
   int a = 1;
   int b = 2;
    do {int temp = a; a = b; b = temp; } while(0);
   return 1;
}

Type traits

We have some compile time functions to infer properties of types.


_is_char()
The three types char, signed char, and unsigned char are collectively called the character types.

_is_pointer
Pointer to object or function

_is_array
Array type

_is_function
A function type describes a function with specified return type. 

_is_floating_point
float, double, and long double return true

_is_integral
The standard signed integer types and standard unsigned integer types are collectively called the
standard integer types;

_is_arithmetic
Integer and floating types are collectively called arithmetic types. 

_is_scalar
Arithmetic types, pointer types, and the nullptr_t type are collectively called scalar types

Note: Type traits that can be easily created with _Generic will be removed. _

Extension - Object lifetime checks

See ownership

Extension assert built-in

In cake assert is an built-in function. The reason is because it works as tips for flow analysis.

For instance, in a linked list when head is null tail is also null, and tail->next always points to null.

Assertion will check these properties in runtime and also make the static analysis assume that assert evaluates to true.


void list_push_back(struct list* list,
                    struct item* _Owner p_item)
{
   if (list->head == NULL) {
      list->head = p_item;
   }
   else {
      assert(list->tail != nullptr);
      assert(list->tail->next == nullptr);
      list->tail->next = p_item;
   }
   list->tail = p_item;
}

However, assert is not a "blind override command." In situations like:

    int i = 0;
    assert(i != 0);

In situations where static analysis can identify two or more possible states, assert works as a state selector, similar to what happens in if statements but without the scope.

    void f(int * _Opt p)
    {
        if (p != NULL) {
           //p is not null here...
        }
    }
    
    void f2(int * _Opt p)
    {
        assert(p != NULL);
        //we assume p is not null here...        
    }