Re: getopt compatibility

On May 27, 11:35am, [email protected] ("James K. Lowden") wrote:
-- Subject: Re: [freetds] getopt compatibility

| ZIGLIO, Frediano, VF-IT wrote:
| > Testing for HP-UX from code is not that
| > difficult and optreset variable can be tested quite easily in autoconf,
| > something like
| > 
| > static int 
| > reset_getopt(void)
| > {
| > #if hpux
| >   optind = 1;
| > #else
| > #if HAVE_OPTRESET
| >   optreset = 1; 
| > #endif
| >   optind = 0;
| > #endif
| > }
| 
| www.gnu.org/software/libtool/manual/libc/Getopt.htm
| 
| Actually, this is going to be much easier than you think.  
| 
| 1.  tsql appears to be wrong: GNU says "The initial value of this variable
| is 1."  Not zero.  
| 2.  NetBSD, HP/UX, and opengroup.org all say the same thing.  
| 
| I tested getopt(3) on a RHEL system; setting optind to 1 caused it to
| parse again from the beginning.  
| 
| 3.  I think it's legal in C to declare 
| 
| 	extern int optreset;
| 
| without ever having:
| 
| 	int optreset;
| 
| so there's no need even to test HAVE_OPTRESET:
| 
| 	static int 
| 	reset_getopt(void)
| 	{
| 	  optreset = 1; 
| 	  optind = 1;
| 	}
| 
| which of course means we don't need the funciton at all.  :-)

Yes, it is, but then you get a link error unless you make it weak.
setting optind always to 1 is probably the best way. For the belt
and suspenders version, setting optreset if it exists is good too.

christos