Confusing about _times_r implementation and newlibc clock

[jolivepetrus]

I'm a bit confusing about _times_r implementation.

Current implementation:

clock_t IRAM_ATTR _times_r(struct _reent *r, struct tms *ptms)
{
clock_t t = xTaskGetTickCount() * (portTICK_PERIOD_MS * CLK_TCK / 1000);
ptms->tms_cstime = t;
ptms->tms_cutime = 0;
ptms->tms_stime = t;
ptms->tms_utime = 0;
struct timeval tv = {0, 0};
_gettimeofday_r(r, &tv, NULL);
return (clock_t) tv.tv_sec;
}

In newlibc clock implementation is:

clock_t
clock ()
{
struct tms tim_s;
clock_t res;

if ((res = (clock_t) _times_r (_REENT, &tim_s)) != (clock_t) -1)
res = (clock_t) (tim_s.tms_utime + tim_s.tms_stime +
tim_s.tms_cutime + tim_s.tms_cstime);

return res;
}

It seems that ticks returned by clock are multiplied by 2, due to in _times_r tms_cstime and tms_stime are set with the current processor ticks. tms_cstime or tms_stime must be set to 0.

[igrr]

Thanks, I haven't checked clock, will fix and add a test for it.

[igrr]

Okay, i have found why i implemented _times_r incorrectly in the first place. I missed one 's' in this line of manpage related to times function:

tms_cstime  The sum of the tms_stimes and tms_cstimes of the child processes.
                                    ^

[igrr]

Fixed in master