In <http://lists.gnu.org/archive/html/bug-tar/2006-06/msg00023.html>
Denis Excoffier reports that compiling tar-1.15.91 generates the
following bogus error message on powerpc-apple-darwin8.6.0:
openat.c: In function 'rpl_openat':
openat.c:60: warning: 'mode_t' is promoted to 'int' when passed through
'...'
openat.c:60: warning: (so you should pass 'int' not 'mode_t' to 'va_arg')
openat.c:60: note: if this code is reached, the program will abort
I reproduced the problem on my i686-pc-linux-gnu platform, so the
problem is generic. At the end of this description is a sample test
case that illustrates the problem. Compiling it with 'gcc -c' yields
the messages:
t.c: In function 'foo':
t.c:18: warning: 'mode_t' is promoted to 'int' when passed through '...'
t.c:18: warning: (so you should pass 'int' not 'mode_t' to 'va_arg')
t.c:18: note: if this code is reached, the program will abort
This diagnostic is bogus: line 18 cannot possibly be executed because
sizeof (mode_t) < sizeof (int). GCC knows this, since it doesn't
generate code for line 18, so GCC should be smart enough to not
generate the diagnostic for line 18.
#include <stdarg.h>
typedef unsigned short int mode_t;
int
foo (int flags, ...)
{
mode_t mode = 0;
va_list arg;
va_start (arg, flags);
/* If mode_t is narrower than int, use the promoted type (int),
not mode_t. Use sizeof to guess whether mode_t is nerrower;
we don't know of any practical counterexamples. */
if (sizeof (mode_t) < sizeof (int))
mode = va_arg (arg, int);
else
mode = va_arg (arg, mode_t); /* This is line 18. */
va_end (arg);
return flags == mode;
}
--
Summary: bogus warning for va_arg argument promotion
Product: gcc
Version: 4.1.1
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: c
AssignedTo: unassigned at gcc dot gnu dot org
ReportedBy: eggert at gnu dot org
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28106
