[tz] proposed changes for Win32 and a improved mktime() algorithm

[Paul Eggert]

Kees Dekker wrote:
> the patch will already give you an
> impression about its mindset.

Unfortunately the patch doesn't work well in boundary cases like overflow or 
leap seconds, and (more generally) it is duplicating functionality that is 
already present in tzcode. If the patch works for you in your situation that's 
fine, but it's not generally suitable for tzcode.