[tz] POSIX extension check

Paul Eggert eggert at cs.ucla.edu
Mon Sep 9 06:14:51 UTC 2013


Arthur David Olson wrote:
> It's probably worthwhile to check whether extended POSIX in zic-produced
> binaries will cause grief when those binaries are used with old versions of
> localtime.c. If so, one or more of documentation, a zic warning, or a
> version number bump may be in order.

Those suggestions sound good, and I'll soon propose
a patch along those lines.

I checked with the test script appended to the message,
and it looks like we're pretty safe.  The only problem
I observed was with America/Godthab for dates past the
start of 2038; this is because the new POSIX TZ string
uses negative hours, which the old tz code does
not grok.  I doubt whether the Godthab issue is urgent,
since we are talking only far-future dates here, and
our predictions for them are quite possibly wrong anyway.

export LC_ALL=C
o=/tmp/tz-2013dd
n=/tmp/tz

omit_junk='s/^[^[:blank:]]*[[:blank:]]*//; s/ UTC / UT /'

for zone in America/Los_Angeles Antarctica/Palmer Asia/Tehran Asia/Gaza Asia/Hebron Pacific/Fiji America/Godthab America/Argentina/San_Luis America/Santiago Pacific/Easter; do
  echo $zone report
  $o/etc/zdump -v $o/etc/zoneinfo-posix/$zone | sed "$omit_junk" >oo
  $o/etc/zdump -v $n/etc/zoneinfo-posix/$zone | sed "$omit_junk" >on
  $n/etc/zdump -v $o/etc/zoneinfo-posix/$zone | sed "$omit_junk" >no
  $n/etc/zdump -v $n/etc/zoneinfo-posix/$zone | sed "$omit_junk" >nn
  set oo on no nn
  for i in oo on no; do
    case $i in
      oo) js='on no nn';;
      on) js='no nn';;
      no) js='nn';;
    esac
    for j in $js; do
      diff -u $i $j
    done
  done
done



More information about the tz mailing list