[tz] Direct implementation of mktime_z

[Ted Phelps]

Hi Guy,

On 24/9/22 17:27, Guy Harris wrote:
> This appears to be more of "faster implementation of mtkime()" than "direct implementation of mtkime_z()", as the bulk of what you're talking about appears to be a faster way of converting a struct tm into a time_t rather than adding support for having a routine that can convert a struct tm, in an *arbitrary* tzdb region rather has the current tzdb region, to a time_t.  The two are independent - the "convert a struct tm, representing local time in a given tzdb region, to a time_t" part is orthogonal to the "is the given tzdb region the default region for the process, or is it an arbitrary region specified by a handle?"

That's right, I'm talking about a faster way to convert a struct tm to a 
time_t.  Perhaps "direct" was a poor choice of word; I was trying to 
distinguish between the "binary search the time_t range" approach, which 
is arguably an indirect way to perform the calculation, to an approach 
where the time_t is calculated with a bunch of algebra and whatnot.  I 
can see how my mention of mktime_z could be interpreted to mean what you 
say.  Apologies for the sloppy subject.

> For the latter, one issue si "what's the appropriate time zone name abbreviation for that particular time, and, if there's a way to 'close' a handle for a tzdb region, then, if there's a call that provides a string for that abbreviation, does that string survive closing the handle for the region?"

Indeed, the fact that tm_zone is a pointer is a frustrating part of 
struct tm, but it's not one I'm attempting to address.

-Ted