On 8 May 2018 at 19:27, Peter Maydell <[email protected]> wrote:
> On 8 May 2018 at 18:49, Peter Maydell <[email protected]> wrote:
>> [weird compiler errors]
>
> This fixes them:
>
> --- a/include/qemu/atomic.h
> +++ b/include/qemu/atomic.h
> @@ -187,7 +187,7 @@
> /* Returns the eventual value, failed or not */
> #define atomic_cmpxchg__nocheck(ptr, old, new) ({ \
> typeof_strip_qual(*ptr) _old = (old); \
> - __atomic_compare_exchange_n(ptr, &_old, new, false, \
> + (void)__atomic_compare_exchange_n(ptr, &_old, new, false, \
> __ATOMIC_SEQ_CST, __ATOMIC_SEQ_CST); \
> _old; \
> })
>
> Seems pretty clearly a compiler bug -- rth says newer gcc don't
> do this, so presumably fixed upstream somewhere between gcc 5 and 6.
Standalone testcase, fwiw:
===begin===
/*
* Weirdly, this compiler will complain about FOO(int8_t) but not
* the others:
* $ gcc -g -Wall -Wunused-value -o zz9.o -c zz9.c
* zz9.c: In function ‘foo_int8_t’:
* zz9.c:12:5: warning: value computed is not used [-Wunused-value]
* __atomic_compare_exchange_n(p, exp, des, 0, __ATOMIC_SEQ_CST,
__ATOMIC_SEQ_CST); \
* ^
* /tmp/zz9.c:17:1: note: in expansion of macro ‘FOO’
* FOO(int8_t)
* ^
*
* This is gcc (Ubuntu 5.4.0-6ubuntu1~16.04.9) 5.4.0 20160609
*/
typedef unsigned char uint8_t;
typedef signed char int8_t;
#define FOO(TYPE) \
void foo_##TYPE(TYPE *p, TYPE *exp, TYPE *des) { \
__atomic_compare_exchange_n(p, exp, des, 0, __ATOMIC_SEQ_CST,
__ATOMIC_SEQ_CST); \
}
FOO(int)
FOO(uint8_t)
FOO(int8_t)
===endit===
thanks
-- PMM
