[nSLUG] Re: DST & tzdata
mspencer at tallships.ca
Mon Mar 5 16:04:02 AST 2007
Bill> I use slapt-get and/or swaret to keep my Slackware box mostly
I guess I should find out about those.
Bill> ...installed glibc-zoneinfo-2.3.6-noarch-7_slack11.0, which
Bill> seems to have everything up-to-date:
Ah, good. That turned up as one of 3 options in the online package
Bill> On this subject, can anyone explain why there are
Bill> America/Halifax, Canada/Atlantic, posix/America/Halifax,
Bill> posix/Canada/Atlantic, right/America/Halifax, and
tzcode2007c.tar.gz is the companion to the NIH tzdata file and
contains code to make the relevant utilities and libraries. At line
62 ff., the Makefile in tzcode2007c, offers an option of having data
files the reflect POSIX time, right time or both sets of files.
I gather from the "backward" tzdata file that Canada/Atlantic is "old"
and deprecated. Donald wrote:
DT> The DST solution for our zone in Olsen format requires running the
DT> "backward" part of the zones to make links. That is why it
DT> appears to have missing stuff.
so he may have some idea how the "backward" thing works and why the
deprecated zone names are used.
DT> The zone you absolutely need, is whatever your /etc/localtime
DT> links to.
Ah, good. Seemed reasonable but I wasn't at all clear on that point.
Worthy of note if, like me, you're a novice in this area are:
+ The location of the compiled data files is apparently (?)
hard-coded into libc/glibc as /usr/share/zoneinfo and this is not
the same as the default install location for the tzdata package
from NIH (/usr/local/etc/zoneinfo/).
+ zdump -v Bogus/Zone reports out some info for 1901 and 2038
instead of reporting an error.
+ zdump always (if environment var TZDIR is not set) looks in the
canonical place (which it apparently gets from glibc ) for its
argument. A typo in the argument or any other deviation results
in the useless 1901/2038 output.
Michael Spencer Nova Scotia, Canada .~.
mspencer at tallships.ca /( )\
More information about the nSLUG