No subject

Valery Aronov <valery at ind.tansu.com.au> wrote:
> How is it possible to manage current zone information in MT-safe fashion?

From: Paul Eggert <eggert at twinsun.com>
> Write and use a thread-safe version of the tz library.
> 
> This requires changing the interfaces to localtime, gmtime, tzset, etc.
> (localtime_r is not enough.)  One must also make the locale a
> parameter, for strftime at least.  And I'm afraid that this will
> require going into the guts of the C library, which will make it hard
> to do portably.
> 
> On my (long) list of things to do is to make a thread-safe version of
> the glibc version of the tz library.  Once you know you're dealing
> with that particular library, the job becomes tractable.

The non-re-entrancy, and the globalness, of the time functions has 
long bothered me.  With the C locale itself global, there's not much
to be done about that.

The (draft) standard C++ locale is not (necessarily) global, and 
supports user extensions such as time zone handling.  Now would be
a good time to start defining a standard time zone facet for the C++ 
locale, so that it can be made official at the next round, in a few 
years.

The glibc locale functions being worked on by Ulrich Drepper are 
designed to support a re-entrant (C) interface.  We should co-ordinate
with him to get the time zone code integrated into his architecture.

Please, let's get the interface right this time.

Nathan Myers
ncm at cantrip.org