RTC getEpoch race condition

[ant32t]

Found a race condition for STM32RTC getEpoch

https://github.com/stm32duino/STM32RTC/blob/a3036e0e9605ddc0e13c7cc259082bfe394b6e77/src/STM32RTC.cpp#L1110

time_t STM32RTC::getEpoch(uint32_t *subSeconds)
{
  struct tm tm;

  syncDate();
  syncTime(); // <-- date may have changed and time sometimes reads 00:00:00 of the following date.

[ant32t]

a fix might just check if the hour is 0 ( maybe minute and second too) and call syncDate again

[uzi18]

@ant32 right, so maybe this check should be in syncTime() function

[ant32]

@uzi18 I'm not sure how that would work. From what I can see the other places where syncTime is used only the time is important and other places where syncDate is called only the date is important.

Actually it is also used in the begin() function which could result in a race condition there possibly? It doesn't look to me like something that needs an exact date/time. @fpistm might know if it racing there is an issue.

[fpistm]

Hi @ant32

Thanks for the issue and PR.
I guess the begin should also be updated as well and order also like in getEpoch to be aligned.

[ant32]

@fpistm added the check to the begin function and changed order to match

[fpistm]

Thanks @ant32t
It is merged.