Re: os.time() and the last second of 1969

[Dirk Jagdmann <doj@...>]

these are some good suggestions, but they don't seem to work. I've 
tested on some systems:

OsX doesn't set errno, and will also calculate values for tm_wday and 
tm_ydat.
An old 32bit Linux doesn't set errno, and will also calculate values for 
tm_wday and tm_ydat.
A modern 64bit Linux doesn't set errno, and will also calculate values 
for tm_wday and tm_ydat.


It seems as if all these implementations are not checking the values of 
the struct tm member variables for a valid range, but simply take these 
integer values and use them in their formula to calculate the timestamp 
value.

I briefly looked into the BSD and Linux implementations of these functions:
https://github.com/lattera/freebsd/blob/master/contrib/ntp/libntp/mktime.c
https://github.com/lattera/glibc/blob/master/time/mktime.c

and there is a lot going on. But as said these functions are trying to 
"make up" for values in the struct tm outside the typical range for 
seconds, minutes, months, etc. So it seems that checking for errno or 
unchanged values of tm_yday is not working.


--
---> Dirk Jagdmann
----> http://cubic.org/~doj
-----> http://llg.cubic.org