You just issued a "fix" to timetool for RH 4.2 and 5.2 (I have not seen the RH 6.0 version) to correctly note that 2000 is a leap year. But the program continues to be wrong for "100" years. It will work for now (the next 100 years) because you hard-coded an exception for 2000. Why not do it right and be done with it? The simple rule is leap year if divisible by 4 but not divisible by 100 unless also divisible by 400. That's all. Not hard.
*** Bug 3182 has been marked as a duplicate of this bug. *** Sigh. I thought that one of the main differences between proprietary software and open source software was getting the job done vs. getting the job done *right*. I downloaded the new timetool-2.5-3 rpm, and took a look at how it was patched to fix the year-2000-is-a-leap-year problem. I was dismayed at the sloppiness. While I am not familiar with the scripting language it is written in [wish], I can see that the patch, as well as the original version, display a basic misunderstanding of the Gregorian calendar. In particular, both the old and new versions of the program incorrectly determine the leap status of years such as 1900 and 2100. They are *not* leap years. The correct rule is: Years not divisible by 4 are not leap years. Years divisible by 4 but not 100 are leap years. Years divisible by 100 but not 400 are not leap years. Years divisible by 400 are leap years. There are many sources for this information. For example, on a Red Hat distribution with gcal installed, see /usr/doc/gcal*/doc. Sloppy patches like the one you provided are the kinds of things that created the Y2K mess in the first place. ### Here's the relevant code in /usr/bin/timetool from timetool-2.5-3: proc days_in_month {} { global cl_month cl_year month_length if {$cl_month == 2} { if {[expr $cl_year / 4.0] != [expr $cl_year / 4]} { return 28 } if {[expr $cl_year / 400.0] == [expr $cl_year / 400]} { if {$cl_year == 2000} { return 29 } else { return 28 } } return 29 } else { return $month_length($cl_month) } } ### Here's a better (but untested) solution: proc days_in_month {} { global cl_month cl_year month_length if {$cl_month == 2} { if {[expr $cl_year / 4.0] != [expr $cl_year / 4]} { return 28 } if {[expr $cl_year / 100.0] != [expr $cl_year / 100]} { return 29 } if {[expr $cl_year / 400.0] != [expr $cl_year / 400]} { return 28 } return 29 } else { return $month_length($cl_month) } }
Confirmed in timetool-2.5-5 also. This fix should actually work quite well. ------- Additional Comments From 06/07/99 17:17 ------- timetool is dead. If it is in use 100 years from now, I'll eat my hat. In any case, there will be a completely new time configuration utility available in the next version of Red Hat that will have things "done right". If you are concerned about doing things "right" with timetool itself, you are right -- it is open source -- and you can change it. Those who have complained that the fix is not 100% correct are also right. However, it seems senseless to issue another update at this point when the main problem is solved and a better solution is in the pipeline. ------- Email Received From Andy <andylev> 06/11/99 03:26 -------
OK I don't want people to go home unhappy. I'm going to slip-stream a fixed version of timetool onto the updates ftp site, and update the errata, but I don't think it warrants sending another announcement and forcing tons of people who don't really need or care about the "correct" fix to get it. timetool-2.6 will be the version.