Mail Archives: djgpp/2000/04/13/18:20:41
Steamer wrote:
>
> I didn't expect that - all the 2.03 zoneinfo files appear to broken.
Yes, they are.
The interesting thing is that if you build them on your machine from
djtzs203.zip, the produced files work correctly. So the problem is evidently
in how the timezone files were compiled on DJ Delorie's machine when the
v2.03 distribution was made. This was done by cross-compiling on a Unix box,
and it seems something goes awry when the timezone files are produced in that
cross environment.
That's why I didn't initially see the problem: I used timezone files I
produced myself on my machine.
And it's not the big/little-endian schizophrenia, either (TZ files are
endian-independent). Hmmm...
> I suppose no-one noticed this before because DJGPP 2.03 was only
> released in January, and the files probably worked OK before the change
> to daylight saving time.
Nope, the v2.03 files are just plain wrong, period. To verify, set the date
to something before DST and rerun the test program.
So I suspect that the fact that nobody noticed this speaks volumes about how
much these files are used and how many people pay attention to the related
problems... ;-)
> As far as I can tell, almost every file in djtzn203.zip is wrong.
> I think it ought to be withdrawn.
Why withdraw it if we can fix it? ;-)
I've just uploaded a new djtzn203.zip to SimTel.NET (it will take a day or
two to propagate to all the mirrors). They were created on my DJGPP machine,
and were tested to produce the expected results. I also took this
opportunity to update the sources of the time-zone info from the latest
version of the data base on elsie.nci.nih.gov.
People who run TZ-sensitive programs and/or pass file archives between
different time zones, are encouraged to upgrade.
Thanks to everyone who contributed to the solution of this bug!
- Raw text -