time_util wrong localtime output (2021/02/29)?

[APochiero]

Hi, 

I'm currently working on a ble mesh project with 2 nodes using nrf Connect SDK v1.4.2 with toolchain 1.4.2:

• Device A with Time Client Model - nrf52840
• Device B with Time Server - nrf5340 

Dev.A sends the following time status to Dev.B in order to update its time:

struct bt_mesh_time_status status = {
      .tai = {
        .sec =  667930834 + k_uptime_get() / 1000,
        .subsec = 1,
      },
      .uncertainty = 0,
      .tai_utc_delta = 37,
      .time_zone_offset = 0,
      .is_authority = true,
};

Those specific TAI seconds indicate a time in the afternoon of 2021/03/01, but bt_mesh_time_srv_localtime's output called on Dev.B returns 

a struct tm with the following values:

 

that is 2021/02/29. 

So I looked at the code using the debugger:

The tai_to_ts function of time_util.c correctly calculates 2021 as a non-leap year, but no check is made on the increment at line 107.

I ran another test by adding 1 day (86400 seconds) to the above time status and the localtime returns 2021/03/02, as if March 1 did not exist.

Is this a bug? 

Adding the following check at the same line could solve the problem:

    if (day_cnt + 1 > months[timeptr->tm_mon]) {
      timeptr->tm_mday = 1;
      timeptr->tm_mon++;
    } else 
      timeptr->tm_mday = day_cnt + 1;

 Regards,

A. Pochiero

[Joakim Jakobsen]

Hi.

I'll forward this to our Mesh developers so that they can comment.

Br,
Joakim

[Joakim Jakobsen]

There has been made improvements to the latest release of the NCS.

Do you still see this using the latest release?

Br,
Joakim

[APochiero]

We are currently working on a project using NCS v1.4.2 with zephyr 2.4 and we don't have plan to update our SDK for now.

However, looking at the new version here, I guess that problem may still occur since no checks are done on that increment at line 89.

Br,

AP