The RTC driver topic

[Giovanni]

Post the code and show me where it is too complex, we will find a solution.

I imagine that if you have a conversion function xyz(native, final) you can have multiple implementations:

Code: Select all

#if defined(STM32F1)
xyz(native, final) {
}
#elif defined(STM32F2_4)
xyz(native, final) {
}
#elif
....

It does not look complex just the code is into another module.

Giovanni

[barthess]

It is not too complex. It just lays in slightly unexpected place. Any way, rev. 4020 contains finished ideas.

[Giovanni]

Barthess, when you have a little time could you review the ticket 3567597?

https://sourceforge.net/tracker/?func=d ... tid=995489

thanks,
Giovanni

[barthess]

I have reviewed the ticket. Problem confirmed. I have merged fix to trunk and test it on STM32-103STK board.

[Giovanni]

Thanks.

Giovanni

[Giovanni]

Hi,

Barthess, there is another one, this time regarding RTCv2, could you examine it? ticket 3568626.

https://sourceforge.net/tracker/?func=d ... tid=995489

thanks,
Giovanni

[barthess]

Accepted. I am working on it.

[barthess]

Giovanni,
After investigating of bug i have found following:
1) this is serous bug able to cause unexpected issues and crushes.
2) localtime() returns pointer to statically allocated buffer with calculated result. In other
words, the function is not thread safe. I have changed it to localtime_r() analog.
Fixed in rev. 4694
UPDATE Tested on my custom board with F4x. I am not able to test it in F1x, could
any body test it on F1x?

[Giovanni]

Thanks,

I am worried about portability, localtime_r() is specific of newlib and we support other compilers. Probably we should study a solution with FAT time directly supported in the driver without any external dependency.

Giovanni

[barthess]

Create timestamp on device with hardware calendar is not a big problem, but for other devices we must create time conversion function. This does not look trivial for me. Probably some one has his own realization.