bug#22302: 25.1.50; time-stamp ignores time-stamp-time-zone

[Eli Zaretskii <eliz@gnu.org>]

> Cc: rgm@gnu.org, 22302@debbugs.gnu.org, hanche@math.ntnu.no
> From: Paul Eggert <eggert@cs.ucla.edu>
> Date: Tue, 12 Jan 2016 13:29:23 -0800
> 
> On 01/12/2016 12:51 PM, Eli Zaretskii wrote:
> >>   don't know what MS-Windows does with TZ="UTC"
> > Nothing.  It doesn't understand that.  The MS-Windows time routines
> > need an explicit offset from UTC to be able to account for the
> > timezone.  So unless you say UTC-0, nothing will happen.
> 
> In that case I'm a bit puzzled. Before today, lines like this:
> 
> # time-stamp-time-zone: "UTC"
> 
> appeared in files like build-aux/gitlog-to-changelog. Emacs implements 
> these lines by temporarily setting the TZ environment variable to "UTC". 
> If that setting is ineffective on MS-Windows, presumably any relevant 
> time stamps were generated in local time instead, which is not wanted.
> 
> Perhaps no MS-Windows users noticed, or none of them bothered to file a 
> bug report.
> 
> Or perhaps "nothing will happen" means "no offset from UTC will be 
> applied", which means that unrecognized TZ settings act like UTC; this 
> is what tzcode does.

I see that what I said was inaccurate: the MS-Windows runtime indeed
doesn't interpret the 3-letter TZ name, but if the offset is missing,
it is taken as zero.  So "UTC" will indeed be interpreted as "UTC+0"
(and so will "EDT" and any other "XXX" with no offset).

Of course, currently this doesn't work in time-stamp, for the reasons
described in the original bug report.  But format-time-string does
behave on Windows as described above.

Sorry for any confusion I could cause.