Re: [PATCH] posixtm: don't reject a time with "60" as the number of seconds

[Jim Meyering]

Paolo Bonzini wrote:
>>      --- -   2009-09-14 14:32:58.934253776 +0200
>>      +++ in  2009-09-14 14:32:58.930840392 +0200
>>      @@ -1,6 +1,6 @@
>>      -12131415.16     13   1260713716 Sun Dec 13 14:15:16 2009
>>      -12131415.16     13   1260713716 Sun Dec 13 14:15:16 2009
>>      -000001010000.00 13 -62167219200 Sat Jan  1 00:00:00 0
>>      +12131415.16     13   1039788916 Fri Dec 13 14:15:16 2002
>>      +12131415.16     13   1039788916 Fri Dec 13 14:15:16 2002
>>      +000001010000.00 13 -62167132800 Sun Jan  1 00:00:00 0000
>>       190112132045.52 13  -2147483648 Fri Dec 13 20:45:52 1901
>>       190112132045.53 13  -2147483647 Fri Dec 13 20:45:53 1901
>>       190112132046.52 13  -2147483588 Fri Dec 13 20:46:52 1901
>>
>> Note that the first two differences are to be expected.
>> Those tests pass only if 2002 is the current year.
>> However, I don't know about the difference in the 3rd line.
>> Note the s/Sat/Sun/ change.  The new number of seconds is 86400
>> larger than the old one; that's exactly one day's worth of seconds,
>> and hence in line with the Sat/Sun change.
>>
>> I don't yet know if this change is a problem in mktime or due
>> to some intervening fix.
>
> cal accepts only years from 1 to 9999, but anyway it implements the
> Julian calendar correctly so it is not useful (you really should try
> "cal 9 1752"!!).

Ouch ;-)

> Going backwards from "cal 1 1" you can see that in
> the Julian calendar 01-Jan-0000 was a Thursday, but that's not so
> relevant.
>
> However cal can help seeing that 01-Jan-0000 is a Saturday in
> Gregorian proleptic calendar (i.e. extending Gregorian calendar before
> the day when it was adopted).  400 years have 146097 days, which is
> divisible by 7, and 01-Jan-2000 was a Saturday.

If that's true, then this new test failure suggests there's a bug in mktime.