[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: 19991229: y2k bug detected in ldm5.0.8, pqact (palt.c) (fwd)


===============================================================================
Robb Kambic                                Unidata Program Center
Software Engineer III                      Univ. Corp for Atmospheric Research
address@hidden             WWW: http://www.unidata.ucar.edu/
===============================================================================

---------- Forwarded message ----------
Date: Wed, 29 Dec 1999 16:21:26 -0700
From: Russ Rew <address@hidden>
To: Albion Taylor <address@hidden>
     address@hidden, address@hidden, address@hidden,
     address@hidden, address@hidden, address@hidden,
     address@hidden, address@hidden, address@hidden
Subject: Re: 19991229: y2k bug detected in ldm5.0.8, pqact (palt.c) 

>To: <address@hidden>
>From: "Albion Taylor" <address@hidden>
>Subject: y2k bug detected in ldm5.0.8, pqact (palt.c)
>Organization: .
>Keywords: 199912292224.PAA23381

Hi Albion,

> We have been running ldm for quite some time, on an AIX 4.2.1 platform.  Our
> pqact.conf file contains lines similar to
> 
> WMO     ^S[AP][AC-Z][A-Z].* .... ([0-3][0-9])
>         FILE    data/obs/surface/hrly/(\1:yy)(\1:mm)\1.hrly
> 
> Some upstream products had time indicators of 01, indicating Jan 1 a little
> early.  These were unexpectedly placed into a file named
> 
> -rw-r--r--   1 weather  staff        582 Dec 29 14:30 100101.hrly
> 
> as if the date information were expanded in a two-digit year of 10 (2010)
> and a month of 01 and a day of 01.
> 
> An examination of the code in palt.c (around lines 641 to 645) suggests
> that, if struct tm has a time_s->tm_year value which is year -1900 (i.e. 100
> for 2000, 101 for 2001, etc), then (\n:yy) in pqact.conf files will expand
> to the first two digits of a number greater than 100; i.e. we will have a
> "year" value of 10 for the years 2000 to 2009, 11 for 2010 to 2019, etc.
> 
> I would suggest some attention be given to debugging this code.

Yes you are exactly right, and here's a patch

$ diff palt.c~ palt.c
643c643
<               (void) sprintf(ostring,"%d",year);
---
>               (void) sprintf(ostring,"%02d", year % 100);

It turns out there is a less serious bug that would cause problems for
dates before 1986 a few lines earlier, that is fixed by the patch:

637c637
<               year = time_s->tm_year + (time_s->tm_year > 87 ? 1900 : 2000);
---
>               year = time_s->tm_year + 1900;

We're very grateful you pointed this out, and we'll be sending an
announcement to the ldm-users and idd mailing lists as soon as we get
new binaries built that contain this fix, so people can download the
binaries instead of building from source.  It looks like only the
pqact executable is affected, so we'll make that executable available
for various platforms to make it easier to install the fix.

I'm a little embarrassed we didn't uncover this one in testing, but we
apparently used the (\1:yyyy) pattern in our testing and didnt't check
the (\1:yy) pattern.  Anyway, thanks again!

--Russ

_____________________________________________________________________

Russ Rew                                         UCAR Unidata Program
address@hidden                     http://www.unidata.ucar.edu