I have found the system specific definition and recommended usage practice of main() hardly documented in context of POSIX getopt() in OSX/Darwin, such as shown in this little test I wrote:
#include <stdio.h>
#include <stdlib.h>
#include <unistd.h>
int main (int argc, const char *argv[])//this is *not* MY "prototype" of main - it is OSX!
{
int opt = 0;
int ch;
char *cvalue = NULL;
while ((ch = getopt (argc, argv, "d:")) != -1)
switch (ch){
case 'd':
cvalue = optarg;
opt = 1;
break;
case '?':
if (optopt == 'd')
printf ("Option -%c requires an argument.\n", optopt);
break;
default:
abort();
}
if(opt)printf("optarg = %s\n", optarg);
return 0;
}
The compilers issue the following, rather intimidating warnings:
gcc:
test.c: In function ‘main’:
test.c:15: warning: passing argument 2 of ‘getopt’ from incompatible pointer type
clang:
test.c:15:32: warning: passing 'const char **' to parameter of type 'char *const *'
discards qualifiers in nested pointer types [-Wincompatible-pointer-types]
while ((ch = getopt (argc, argv, "d:")) != -1)
^~~~
/usr/include/unistd.h:548:31: note: passing argument to parameter here
int getopt(int, char * const [], const char *) __DARWIN_ALIAS(getopt);
^
AFAIK, char pointers to immutable arrays are indeed incompatible with immutable pointers to
character arrays. There is not much I can do unless changing the main() definition to
supply arguments as required by POSIX getopt():
int main(int argc, char *const argv [] ){ ... }
or according to the C standard (as defined in section 5.1.2.2.1 of C11 draft standard n1570) and described in this SO post:
int main(int argc, char *argv[] ){ ... }
This silences the compiler warnings, but makes me wonder: although same manufacturer implements both C and POSIX, forking from their prototype of main() which seemingly conflicts with their hosted environment implementation of POSIX C gives an impression of solving the problem.
As I am developing a set of UNIX command line tools in OSX, all of which must reliably redirect, pipe, stay interactive, use files
etc. in rather complex chains and critical conditions, I need a bit deeper understanding for what reason the OSX main() should take such argument types. Has someone worked with this specific implementation of main(), and which would the safe procedure of solving the described issue in OSX/Darwin, without inducing risks of UB or otherwise compromised functionality? This is as far as my efforts have brought me - either in conflict with the system, or with the C standard. Thanks in advance.
