[Linux-HA] Problems with Year 2007?

[Christian Froemmel]

On Mon, 1 Jan 2007, Marian Neubert wrote:

Hello,

> crm_verify[11179]: 2007/01/01_12:13:26 ERROR: 
> ordinal_to_gregorian:iso8601.c Triggered non-fatal assert at 
> iso8601.c:618 : a_date->has->yeardays

I found at least for this assert the error in the code.

In file "lib/crm/common/iso8601.c" starting on line 827 there is a check
for "rhs->tm_yday > 0":

        if(rhs->tm_yday > 0) {
                /* days since January 1 [0-365] */
                lhs->yeardays = 1 + rhs->tm_yday;
                lhs->has->yeardays =TRUE;
        }

The comment inside the block is absolutely correct, so the first day of
year is 0 in tm-structs - and the block is never executed on every new
year of every year. I think it should read:

        if(rhs->tm_yday >= 0) {
                /* days since January 1 [0-365] */
                lhs->yeardays = 1 + rhs->tm_yday;
                lhs->has->yeardays =TRUE;
        }

> crm_verify[11179]: 2007/01/01_12:13:26 ERROR: 
> convert_from_ordinal:iso8601.c Triggered non-fatal assert at 
> iso8601.c:609 : ordinal_to_gregorian(a_date)
> crm_verify[11179]: 2007/01/01_12:13:26 ERROR: ha_set_tm_time:iso8601.c 
> Triggered non-fatal assert at iso8601.c:849 : rhs->tm_mon < 0 || 
> lhs->months == (1 + rhs->tm_mon)

This think those asserts are side-effects of the above error, since the
function "ordinal_to_gregorian()" fails in the first assert.. But I am not
100% sure.

But the 100%-solution is: don't work on new years day ;-)

Regards,
Christian
-- 
SYSTEM CONCEPT GmbH
Systemadministration/Netzwerk