ede/ede-bug-tools/ede-bug-report/xmlrpc-c/deps/util/include/stdargx.h
2009-07-03 14:23:51 +00:00

68 lines
1.9 KiB
C

#ifndef STDARGX_H_INCLUDED
#define STDARGX_H_INCLUDED
#include "xmlrpc_config.h"
#include <stdarg.h>
#include <string.h>
/*----------------------------------------------------------------------------
We need a special version of va_list in order to pass around the
variable argument heap by reference, thus allowing a subroutine to
advance the heap's pointer.
On some systems (e.g. Gcc for PPC or AMD64), va_list is an array.
That invites the scourge of array-to-pointer degeneration if you try
to take its address. Burying it inside a struct as we do with out
va_listx type makes it immune.
Example of what would happen if we used va_list instead of va_listx,
on a system where va_list is an array:
void sub2(va_list * argsP) [
...
}
void sub1(va_list args) {
sub2(&args);
}
This doesn't work. '&args' is the same thing as 'args', so is
va_list, not va_list *. The compiler will even warn you about the
pointer type mismatch.
To use va_listx:
void sub1_va(char * format, va_list args) {
va_listx argsx;
init_va_listx(&argsx, args);
sub2(format, &argsx);
}
-----------------------------------------------------------------------------*/
typedef struct {
/*----------------------------------------------------------------------------
Same thing as va_list, but in a form that works everywhere. See above.
-----------------------------------------------------------------------------*/
va_list v;
} va_listx;
static __inline__ void
init_va_listx(va_listx * const argsxP,
va_list const args) {
#if VA_LIST_IS_ARRAY
/* 'args' is NOT a va_list. It is a pointer to the first element of a
'va_list', which is the same address as a pointer to the va_list
itself.
*/
memcpy(&argsxP->v, args, sizeof(argsxP->v));
#else
argsxP->v = args;
#endif
}
#endif