2.3.3 saving to postgresql causes zombie scheduled transactions

[Phil Longstaff]

On August 12, 2009 04:53:29 pm Graham Menhennitt wrote:
> I found (one of) the corruptions that I mentioned in my email of yesterday!
>
> I have an existing GC XML file that contains a scheduled transaction
> that had "repeats until" selected with a date of 28/02/08 (I'm using
> dd/mm/yy format). After saving to postgresql and restarting GC, the same
> transaction now has "forever" selected. "Since last run" now contains
> all the occurances of this transaction from 28/2/08 until now.
>
> I was running with --debug and the trace file is full of:
>
> * 06:38:47  WARN <qof.engine> [gnc_iso8601_to_timespec_gmt()]  mktime
> failed to handle daylight saving: tm_hour=0 tm_year=70 tm_min=0 tm_sec=0
> tm_isdst=-1 for string=1970-01-01 00:00:00
> * 06:38:47  CRIT <qof.engine> [gnc_iso8601_to_timespec_gmt()]  unable to
> recover from buggy mktime
>
> Note that I'm in Australia - a timezone that's ahead of GMT and has
> daylight saving over the new year transition. That might be relevant.
>
> So it looks like there is a time conversion problem when saving to
> postgres. It causes scheduled transactions to come back from the dead.
>
> Should I put it into bugzilla?

Yes.  Please log to bugzilla.

Phil