Re: possible patch for coreutils 5.1.1 failure on OSF 3.2C

[Jim Meyering]

Paul Eggert <address@hidden> wrote:
> address@hidden (Paul Jarc) writes:
>
>> This is still not quite right on systems with TAI clocks.
>
> Good point.  Also, I notice that only the first part of my previous
> patch was installed; presumably this is a typo?  Anyway, here's a
> patch to the latest CVS snapshot that I have.

Thanks to both of you Pauls :-)
I've sort of punted on this one:

        * tests/touch/relative: Test only year/month/day, not hours/min/sec,
        so as to avoid problems with systems using TAI clocks.
        Although it's no longer necessary, set TZ=UTC0 also for the
        initial touch command.  Reported by Paul Jarc here:
        http://article.gmane.org/gmane.comp.gnu.core-utils.bugs/1504

Index: relative
===================================================================
RCS file: /fetish/cu/tests/touch/relative,v
retrieving revision 1.2
retrieving revision 1.5
diff -u -p -r1.2 -r1.5
--- relative    19 Jan 2004 16:26:38 -0000      1.2
+++ relative    21 Jan 2004 19:57:35 -0000      1.5
@@ -17,7 +17,7 @@ trap '(exit $?); exit $?' 1 2 13 15
 framework_failure=0
 mkdir -p $tmp || framework_failure=1
 cd $tmp || framework_failure=1
-touch --date='2004-01-16 05:26:38 +0' f || framework_failure=1
+TZ=UTC0 touch --date='2004-01-16 12:00 +0000' f || framework_failure=1
 
 if test $framework_failure = 1; then
   echo "$0: failure in testing framework" 1>&2
@@ -26,14 +26,14 @@ fi
 
 fail=0
 
-# Set times back by 30 minutes.
-touch --ref f --date='-30 min' f || fail=1
+# Set times back by 5 days.
+touch --ref f --date='-5 days' f || fail=1
 
-TZ=utc command ls -og --time-style=+%Y-%m-%dT%H:%M:%S f > out.1 || fail
+TZ=UTC0 command ls -og --time-style=+%Y-%m-%d f > out.1 || fail
 sed 's/ f$//;s/.* //' out.1 > out
 
 cat <<\EOF > exp || fail=1
-2004-01-16T04:56:38
+2004-01-11
 EOF
 
 cmp out exp || fail=1