The application is not time-crtical, but why insert a bug when you can avoid it. [emphasis added]
The tone of this sentence implies that avoidance of this particular bug is essentially cosmetic and more or less optional.
It's not just theoretically possible; it will eventually happen that successive calls to localtime occur on successive days. This will produce impossible dates (e.g., 31 April) or dates that are 'off' by a month or as much as a year (e.g., 31 Dec 2012 vice 31 Dec 2011).
But if the application can tolerate dates that are impossible or wildly inaccurate, why bother with dates at all?
The rant is valid - thank you for your thoughts. However
This program is started manually, generates code templates. In this special case it is highly unlikely that localtime calls will happen at midnight and/or New Year's Eve.
The dates are only used in commented headers, wasn't my choice :) HOwever it's possible, that sometime somebody will have and idea to read the headers and produce some statistics, so one should take care of it.
Thank you again. Much can be learned from the rants of the experienced :).