James Carlson wrote:
> Don Cragun writes:
>> void
>> timeradd(struct timeval *a, struct timeval *b, struct timeval *res);
> [...]
>> int
>> timercmp(struct timeval *a, struct timeval *b, CMP);
>>
>> int
>> timerisset(struct timeval *tvp);
>>
>> void
>> timersub(struct timeval *a, struct timeval *b, struct timeval *res);
>
> It'd be nice to be able to make 'a', 'b', and 'tvp' const.
>
> (Man, that "CMP" thing sure is evil ...)
>
>> These macros are not available in function form. All of these macros
>> evaulate their arguments more than once. If parameters passed to these
>> macros are expressions with side effects, the results are undefined.
>
> Except for timercmp, where the issue is obvious, it's unclear to me
> why the others must be defined as function-like macros, especially as
> ones with hidden side-effect warts.
>
#define timerclear(t) { struct timeval *a = (t); a->tv_sec = a->tv_usec
= 0; }
seems to be a reasonable fix.
>> lint will complain about a constant conditional for timeradd() and
>> timersub(). Use of the /* LINTED */ comment before timeradd() or
>> timersub() is recommended.
>
> Ick. Is there no way to fix that rather than just documenting it away?
>
As above:
#define timeradd(i1, i2, o1) \
{ struct timeval *a = i1, *b = i2, *result = o1; \
result->tv_sec = a->tv_sec + b->tv_sec; \
result->tv_usec = a->tv_usec + b->tv_usec; \
if (result->tv_usec >= 1000000) { \
++result->tv_sec; \
result->tv_usec -= 1000000; \
} \
}
- Bart
--
Bart Smaalders Solaris Kernel Performance
barts at cyber.eng.sun.com http://blogs.sun.com/barts
"You will contribute more with mercurial than with thunderbird."
