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

[Kees Dekker]

>And even if it's a viable compiler, if the compiler is merely generating 
>false-alarm warnings then we typically do not bother "fixing" the code. Such 
>"fixes" are often more trouble than they're worth. Fixing real bugs takes 
>precedence over pacifying foolish compilers. You can ignore the false alarms 
>with "grep -v".

I forgot something: warnings about not initialized variables are almost always triggered because there is a
If else[/if else].. codepath, where one of the codepaths uses a variable that is not initialized somewhere. Although
This situation may not happen in practice; such warnings point to (IMO) imperfect code. Based on my experience,
such warnings are usually worth to be solved. Maintaining code will not suffer from one initialization at declaration level,
and performance will (usually) also not suffer, as most initialization code just takes one CPU instruction.

Anyhow, I will try to check your patches Friday or otherwise hopefully next week.