On Wed, Jan 27, 2010 at 10:24:08AM -0600, Gunnar Wolf wrote:
> In any case, the potentially offending code (that means, the code that
> generates the date string - For the full code, /usr/bin/swaks lines
> 1532-1552) is:
>
> sub get_date_string {
> (…)
> my @l = localtime();
> (…)
> my @g = gmtime();
> $o = (timelocal(@l) - timelocal(@g))/36;
> (…)
> }
> --- /usr/share/perl/5.10.1/Time/Local.pm 2009-11-21 13:30:05.000000000
> -0600
> +++ /tmp/Local.pm 2010-01-27 10:21:17.000000000
> -0600
> @@ -177,7 +177,11 @@
> my ( $s, $m, $h ) = localtime($loc_t);
> $loc_t -= $dst_off if $s != $_[0] || $m != $_[1] || $h != $_[2];
>
> - return $loc_t;
> + # Comparing two very-slightly-off readings of the clock yields a
> + # difference of -0.027777 seconds. Timezones are integers, and
> + # represent the number of seconds; round (don't truncate) the
> + # result as values ending with '99' are not valid.
> + return int($loc_t + 0.5);
> }
Sorry if I'm dense, but I don't see Time::Local doing anything wrong here.
It's not a case of 0.027777 seconds - both localtime and gmtime have a
resolution of 1 second (as long as Time::HiRes is out of the picture).
Rather, the buggy code in swaks ends up dividing a one second difference
between the localtime() and gmtime() invocations by 36, giving the
magic 0.027777.
The zone handling code in timelocal() already does the right thing: it's
calling localtime with an argument so it doesn't look at the current time.
So should swaks.
perl -e '$|=1;use Time::Local; $sys_tz=`date +%z`*1; $i=0; while (1) {
$ref=localtime; @l=localtime($ref); @g=gmtime($ref); $o=(timelocal(@l) -
timelocal(@g))/36; print "." unless $i++%1000; print $o if $o ne $sys_tz}'
Please reassign back if you agree.
--
Niko Tyni [email protected]
--
To UNSUBSCRIBE, email to [email protected]
with a subject of "unsubscribe". Trouble? Contact [email protected]
