bin/53899: mktime gives wrong result in Central timezone
David MCNETT
nugget at slacker.com
Wed Dec 1 10:40:25 PST 2004
The following reply was made to PR bin/53899; it has been noted by GNATS.
From: David MCNETT <nugget at slacker.com>
To: freebsd-gnats-submit at FreeBSD.org, bp at barryp.org
Cc: stefanf at freebsd.org
Subject: Re: bin/53899: mktime gives wrong result in Central timezone
Date: Wed, 1 Dec 2004 12:34:05 -0600
Barry did all the heavy lifting with this issue, I just wanted to clear
up any remaining confusion.
In researching this issue further I learned that the bug cannot be
reproduced on a 6.0-CURRENT or RELENG_5 machine. The behavior still
exists on any 4.x machine and on older 5.x machines leading me to
speculate that this problem may have been resolved as a byproduct of
the tzcode2004a import performed by stefanf in June 2004.
4.10-STABLE, 21-Oct-2004 exhibits the bug
5.2.1-RELEASE-p5, 20-Apr-2004 exhibits the bug
5.3-STABLE, 7-Nov-2004 is correct
6.0-RELEASE, 1-Dec-2004 is correct.
I also determined that the problem is not within the zic code. MD5
sums match on the generated /usr/share/zoneinfo/America/Chicago files
when comparing an affected 4.x machine to a 6-CURRENT machine which
does not exhibit the behavior. If zic on both systems is producing the
same compiled zone files then it would appear that the problem is in
libc when parsing the data.
It would be nice if the tzcode2004a import, specifically the libc bits,
could be MFC'd into RELENG_4 prior to 4.11-RELEASE, especially since
4.11 is expected to be the final 4.x release.
I was never able to determine what the deal was in Chicago on
1-Mar-1936.
More information about the freebsd-bugs
mailing list