Joe Laffey writes:

This has nothing to do with what localtime does. The objective here is to obtain the difference between the local time, that is what localtime gives, and GMT.

Please post what the NetBSD man page actually says. For reference information, here's the Linux version:

... timezone with the difference between Coordinated Universal Time (UTC) and local standard time in seconds ...

The global timezone variable is defined to contain the difference between GMT, and the local standard time. Then, tm_dst is defined as follows:

tm_isdst A flag that indicates whether daylight saving time is in effect at the time described.

The objective here is to take the local time, computed by localtime(), and compute the difference between that, and GMT. Therefore, the time zone difference is computed by taking timezone, and adding one hour if tm_isdst is set.

Those two lines of code cannot be removed, at least on Linux. You should again review localtime's manpage. You're claiming that the 'timezone' variable is always set by the localtime() function, to indicate the difference between the local time given as its input, and GMT. So, what you're saying that if localtime() receives input for when standard time is in effect, at exit timezone will be 21600 (that is, -0600 GMT), but when localtime() receives input for when daylight time is in effect, at exit timezone will be 18000 (that is, -0500 GMT). That doesn't sound right.

So your timezone right now indicates -05h GMT, which is correct for CDT. You are now implying that if you were to run localtime() and provide a time interval when standard time in effect, at exit timezone would be 21600, or -06h GMT.

Try feeding a January time to localtime(), to find out whether timezone will now read 21600. This would mean a NetBSD-specific hack, in which case I'll need the output of config.sub to code a test for it. That would be unfortunate, because that would mean that NetBSD's timezone functions are not thread-safe.