[Evolution-hackers] evolution-data-server amd64 build trouble/fix

["John Stewart" <jpstewar purdue edu>]

I am seeing problems compiling evolution-data-server 1.0.2 on  
FreeBSD/AMD64 5.3 Beta 7.  The problem is a duplicate macro definition in  
libdb/dbinc/mutex.h.  Has anyone else had this problem?  From the looks of  
it any AMD64 OS could trip on this.  The offending code starts on line 731.


/*********************************************************************
 * amd64/gcc assembly.
 *********************************************************************/
#ifdef HAVE_MUTEX_X86_64_GCC_ASSEMBLY
typedef unsigned char tsl_t;

#ifdef LOAD_ACTUAL_MUTEX_CODE
/*
 * For gcc/amd64, 0 is clear, 1 is set.
 */
#define  MUTEX_SET(tsl) ({           \
	register tsl_t *__l = (tsl);          \
	int __r;              \
	asm volatile("mov $1,%%rax; lock; xchgb %1,%%al; xor $1,%%rax"\
		: "=&a" (__r), "=m" (*__l)          \
		: "1" (*__l)            \
		);                \
	__r & 1;              \
})

#define  MUTEX_UNSET(tsl)  (*(tsl) = 0)
#define  MUTEX_INIT(tsl)   MUTEX_UNSET(tsl)
#endif
#endif

/*********************************************************************
 * x86_64/gcc assembly.
 *********************************************************************/
#ifdef HAVE_MUTEX_X86_64_GCC_ASSEMBLY
typedef unsigned char tsl_t;

#ifdef LOAD_ACTUAL_MUTEX_CODE
/*
 * For gcc/x86-64, 0 is clear, 1 is set.
 */
#define	MUTEX_SET(tsl) ({						\
	register tsl_t *__l = (tsl);					\
	int __r;							\
	asm volatile("movq $1,%%rax; lock; xchgb %1,%%al; xorq $1,%%rax"\
	    : "=&a" (__r), "=m" (*__l)					\
	    : "1" (*__l)						\
	    );								\
	__r & 1;							\
})

#define	MUTEX_UNSET(tsl)	(*(tsl) = 0)
#define	MUTEX_INIT(tsl)		MUTEX_UNSET(tsl)
#endif
#endif

The x86-64 code is triggered by the same macro as amd64, so the build  
fails due to the duplicate typedef/macro definition.  Isn't x86-64  
synonymous with amd64?  It seems like the second would be unnecessary.   
Deleting the x86-64 section allows it to compile and function correctly.   
The second doesn't seem needed.  Forgive me if I am wrong, but isn't  
x86-64 the same as amd64?  I am not familiar with x86 assembly so I am not  
sure which assembly code is correct, but using the amd64 MUTEX_SET macro  
(with mov and xor rather than movq and xorq) I haven't had any problems,  
and I finally the addressbook feature in evolution 2 working.