Pelles C forum

Pelles C => Bug reports => Topic started by: Werner on January 19, 2021, 03:04:46 PM

Title: Evaluation of variadic macros broken
Post by: Werner on January 19, 2021, 03:04:46 PM

The C standard (§ 6.10.3 p. 10) accepts variadic macros in 2 different forms:

# define identifier lparen ... ) replacement-list new-line
# define identifier lparen identifier-list , ... ) replacement-list new-line

While the first form works fine with Pelles C, the second form yields: error #1053: Disagreement in number of macro arguments

E. g.:

#include <stdio.h>
#include <string.h>

#define arg_and_ellipsis(arg, ...) \
{\
   if (!strlen(#__VA_ARGS__)) \
      printf("1 argument passed: %s\n", arg); \
   else \
      printf("more than 1 argument passed: %s %s\n", arg, #__VA_ARGS__); \
}

int main(void)
{
   arg_and_ellipsis("arg_1");
   arg_and_ellipsis("arg_1", "arg_2");
}

Title: Re: Evaluation of variadic macros broken
Post by: Prokrust on January 20, 2021, 03:25:05 PM

//The example below works
//The idea is taken here: https://gustedt.wordpress.com/2010/06/08/detect-empty-macro-arguments/

#include <stdio.h>
#define ARGS_COUNT(...)  _ARGS_COUNT3(_ARGS_COUNT_EMPTY(__VA_ARGS__), __VA_ARGS__)
#define _ARGS_COUNT3(_1,...)  _ARGS_COUNT4(_1,__VA_ARGS__)
#define _ARGS_COUNT4(_1,...)  _ARGS_COUNT##_1(__VA_ARGS__)
#define _ARGS_COUNT(...)  _ARGS_COUNT2(__VA_ARGS__,15,14,13,12,11,10,9,8,7,6,5,4,3,2,1,0)
#define _ARGS_COUNT0(...)  _ARGS_COUNT2(__VA_ARGS__,15,14,13,12,11,10,9,8,7,6,5,4,3,2,0,0)
#define _ARGS_COUNT2(_0,_1,_2,_3,_4,_5,_6,_7,_8,_9,_10,_11,_12,_13,_14,_15, ...)  _15
#define _ARGS_COMMA(...)  ,
#define _ARGS_COUNT_EMPTY(...)  _ARGS_COUNT_EMPTY1(__VA_ARGS__,)
#define _ARGS_COUNT_EMPTY1(_1,...)  _ARGS_COUNT_EMPTY2(_ARGS_COMMA##_1())
#define _ARGS_COUNT_EMPTY2(...)  _ARGS_COUNT_EMPTY3(__VA_ARGS__,0,,)
#define _ARGS_COUNT_EMPTY3(_1,_2,_3,...)  _3

int main(int argc, char *argv[]) {
   printf("()=%i\n",ARGS_COUNT());
   printf("(1,2,3)=%i\n",ARGS_COUNT(1,2,3));
   printf("(,)=%i\n",ARGS_COUNT(,));
   printf("(5)=%i\n",ARGS_COUNT(5));
   printf("(6,)=%i\n",ARGS_COUNT(6,));
   return 0;
}

Title: Re: Evaluation of variadic macros broken
Post by: Werner on January 20, 2021, 04:18:17 PM

Thanks, Prokrust, for your suggestion.

In my eyes, this seems to be a workaround a little bit too convoluted.

As far as I understand the C standard, I have met a compiler bug which I just wanted to report.

By the way, the code provided by me works fine with GCC.

Werner

Title: Re: Evaluation of variadic macros broken
Post by: TimoVJL on January 21, 2021, 05:26:32 AM

no problem with clang
first-aid

Code: [Select]

arg_and_ellipsis("arg_1",);
BTW
SrcFileCC2 Add-In for v9, a second opinion (https://forum.pellesc.de/index.php?topic=8107.msg29671#msg29671)

Title: Re: Evaluation of variadic macros broken
Post by: Werner on January 22, 2021, 04:33:10 PM

Thank you very much for your advice, TimoVJL!

This works, but nevertheless I consider the reported behavior a bug--or am I mistaken?

Anyway, there is also no problem with GCC.

Werner

Title: Re: Evaluation of variadic macros broken
Post by: Pelle on January 25, 2021, 09:15:55 PM

Not a bug. You must specify something for __VA_ARGS__ (this has been debated, and could possibly change in some future C standard, but right now this is how it's supposed to work for C).

Title: Re: Evaluation of variadic macros broken
Post by: Werner on January 27, 2021, 05:08:18 PM

After re-reading the standard several times, I must admit that you are right, Pelle.

§ 6.10.3 p. 4, 2nd sentence, and § 6.10.3 p. 12, 2nd sentence, of the C standard arguably confirm this.
Although, in my personal opinion, not only some textbooks and a lot of posts on the Internet, but also the wording of the standard itself might be somewhat confusing.

Anyway, I apologize for the inconvenience caused!

Werner

Title: Re: Evaluation of variadic macros broken
Post by: Pelle on January 29, 2021, 07:30:44 PM

No problem, and I agree this isn't the best part of the C standard...