mktime not accounting for leap year

will Resnick

All,

I am running mktime based on a GPS time I am receiving. I check the tm struct values and they are all correct. When I check the return value in epoch seconds it is off by 1 day. Also, for the tm.year, it is current year - 1970, not 1900. Is anyone else having this issue?

Thanks,

Will

over 13 years ago

0 Archaeologist over 13 years ago

TI__Guru* 84285 points

The TI compiler library uses a different epoch. See http://processors.wiki.ti.com/index.php/Time_and_clock_RTS_Functions#epoch

0 will Resnick over 13 years ago in reply to Archaeologist

Expert 2155 points

So this explains why the value to subtract is 1970 and not 1900. It still doesn't account for why the mktime function does not account for leapyear this year.

Thanks,

Will

0 Archaeologist over 13 years ago in reply to will Resnick

TI__Guru* 84285 points

Could you please post a short test case demonstrating the problem, showing the values you are passing to mktime and the return value?

0 will Resnick over 13 years ago in reply to Archaeologist

Expert 2155 points

void getUbloxUTCTime(unsigned char* utcTimeBuff, unsigned int *time_sec, int *time_mSec, int *time_nSec)

{

struct tm time_struct;

unsigned short year = 0;

time_struct.tm_sec = utcTimeBuff[NAVTIME_SEC_OFFSET];

time_struct.tm_min = utcTimeBuff[NAVTIME_MIN_OFFSET];

time_struct.tm_hour = utcTimeBuff[NAVTIME_HOUR_OFFSET];

time_struct.tm_mday = utcTimeBuff[NAVTIME_DAY_OFFSET];

time_struct.tm_mon = utcTimeBuff[NAVTIME_MONTH_OFFSET] - 1;

memcpy(&year, &utcTimeBuff[NAVTIME_YEAR_OFFSET], sizeof(short));

time_struct.tm_year = year - 1970;

time_struct.tm_isdst = 0;

*time_sec = mktime(&time_struct);

memcpy(time_nSec, &utcTimeBuff[NAVTIME_NANO_OFFSET], sizeof(time_nSec));

*time_mSec = abs(*time_nSec) / 1000000;

}

Input:

tm_sec 33

tm_min 12

tm_hour 13

tm_mday 22

tm_mon 4

tm_year 42

tm_wday 0

tm_yday 0

tm_isdst 0

Output:

time_sec = 1337605953

Converting time_sec gives Mon, 21 May 2012 13:12:33 UTC

0 will Resnick over 13 years ago in reply to will Resnick

Expert 2155 points

I believe TI's mktime does not account for the day once past FEB 29 if we are within the leapyear. Everything was fine until passing FEB 29.

THanks,

Will

0 Archaeologist over 13 years ago in reply to will Resnick

TI__Guru* 84285 points

mktime is messing up leap years because it starts computing them from 1900 because it assumes year 0 is 1900, and your code assumes year 0 is 1970. The tm_year field in mktime uses an epoch of 1900, as required by the C standard, so you can't use 1970 as the bias. You must use 'year - 1900' as the value of tm_year. Then, if you want to convert the return value to the POSIX 1970-biased number of seconds, subtract 2208988800L seconds (the number of seconds from midnight January 1, 1900 to the POSIX epoch, as defined in RFC 868).

Code Composer Studio™︎

Code Composer Studio forum

mktime not accounting for leap year