[STM32XX] Fix RTC minimum date

[svastm]

Problem

• The MBED_16 fail on STM32F0 targets.
• None of the ST targets are able to set_time(0) => year 1970 correctly.

Fix

STM32F0, STM32F3, STM32F4, STM32F7, STM32L0, STM32L1, STM32L4

They use a BCD format to store the date in the RTC. The year is store on 2 * 4 bits. Because the first year is reserved to see if the RTC is init, the supposed range is 01-99. The idea is to cover the standard range from 1970 to 2038 (limited by the 32 bits of time_t).

The fix move the RTC range to match the standard one. The idea is to keep the year 1970 and the leap years synchronized. By moving it 68 years forward from 1970, it become 1969-2067 which include 1970-2038. 68 is also a multiple of 4 so it let the leap year synchronized. Between 1904 and 2096 there is a leap year every 4 years.

STM32F1

They don't use a BCD format but a 32bit register for the seconds and a software structure to store dates. It should not be a problem to not use shifts.

Test

I used the following code to test the fix.

#include "mbed.h"
#include "test_env.h"

#if !DEVICE_RTC
  #error [NOT_SUPPORTED] RTC is not supported
#endif

int main() {
    MBED_HOSTTEST_TIMEOUT(20);
    MBED_HOSTTEST_SELECT(st_rtc_min);
    MBED_HOSTTEST_DESCRIPTION(st_rtc_min);
    MBED_HOSTTEST_START("ST_RTC_MIN");

    struct tm timeinfo;
    time_t seconds = 0;
    char buffer[32] = {0};
    bool result = true;

    // test the minimum date
    // Set time to 1 Jan 1970 00:00:00
    timeinfo.tm_mon  = 0;
    timeinfo.tm_mday = 1;
    timeinfo.tm_year = 70;
    timeinfo.tm_hour = 0;
    timeinfo.tm_min  = 0;
    timeinfo.tm_sec  = 0;
    timeinfo.tm_isdst  = -1;
    set_time(mktime(&timeinfo));

    // Print the time
    seconds = time(NULL);
    strftime(buffer, 32, "%Y-%m-%d %H:%M:%S %p", localtime(&seconds));
    printf("MBED: [%ld] [%s]\r\n", seconds, buffer);

    if (strcmp(buffer,"1970-01-01 00:00:00 AM") != 0) {
      result = false;
    }

    // test the leap years
    // Set time to 28 Feb 2032 23:59:59
    timeinfo.tm_mon  = 1;
    timeinfo.tm_mday = 28;
    timeinfo.tm_year = 132;
    timeinfo.tm_hour = 23;
    timeinfo.tm_min  = 59;
    timeinfo.tm_sec  = 59;
    timeinfo.tm_isdst  = -1;
    set_time(mktime(&timeinfo));
    wait(1.1);

    // Print the time
    seconds = time(NULL);
    strftime(buffer, 32, "%Y-%m-%d %H:%M:%S %p", localtime(&seconds));
    printf("MBED: [%ld] [%s]\r\n", seconds, buffer);

    if (strcmp(buffer, "2032-02-29 00:00:00 AM") != 0) {
      result = false;
    }

    // test the maximum date
    // IAR use to fail on this one because his year 2038 bug seems to 
    // happen the 6 Feb 2036
    // Set time to 19 Jan 2038 03:14:00
    timeinfo.tm_mon  = 0;
    timeinfo.tm_mday = 19;
    timeinfo.tm_year = 138;
    timeinfo.tm_hour = 3;
    timeinfo.tm_min  = 14;
    timeinfo.tm_sec  = 0;
    timeinfo.tm_isdst  = -1;
    set_time(mktime(&timeinfo));
    wait(1.1);

    // Print the time
    seconds = time(NULL);
    strftime(buffer, 32, "%Y-%m-%d %H:%M:%S %p", localtime(&seconds));
    printf("MBED: [%ld] [%s]\r\n", seconds, buffer);

    if (strcmp(buffer, "2038-01-19 03:14:01 AM") != 0) {
      result = false;
    }

    notify_completion(result);
}

Result

Target	toolchain	min date	max date
F0,F3,F4,F7,LX	uARM, ARM, GCC	1 Jan 1969	19 Jan 2038
F0,F3,F4,F7,LX	IAR	1 Jan 1969	6 Feb 2036
F1	uARM, ARM, GCC	13 Dec 1901	19 Jan 2038
F1	IAR	13 Dec 1901	6 Feb 2036

[0xc0170]

@mbed-bot: TEST

HOST_OSES=windows
BUILD_TOOLCHAINS=ARM,GCC_ARM,IAR
TARGETS=ALL

[0xc0170]

Thanks for providing the detailed description, for future, part of it (or even all) can be included in the commit message.

LGTM

[mbed-bot]

[Build 625]
FAILURE: Something went wrong when building and testing.