GnuCash Daily Source Diff

[Linas Vepstas]

On Fri, Oct 19, 2001 at 02:04:34PM -0700, Dave Peticolas was heard to remark:
> > I think that you were trying to fix a bug related to changes to kvp
> > values in multi-user mode.  Is this so?
> 
> It actually turned up in single-user mode. When deleting a split
> in a transactoin, it would try to delete the kvp frame of the split.
> But the original sql command didn't specify the object type for the
> audit record of the kvp frame 

It shouldn't need to, it should have been copied in ... 
The INSERT INTO ... SELECT ... construction is essentially a copy
command.  The objtype is supposed to be copied from what the 
SELECT returned.  Several fields, not just the objtype, are supposed 
to be copied, including the value of 'data' , 'type', 'num', 'denom', 
etc.

> and this
> was causing the query to fail, at least under Postgres 7.0. 

Hmmm. Not good.  I would hate to have to do the following:

#define CPY_KVP(TYPE)                                                   \
{                                                                       \
   p = stpcpy (p, "INSERT INTO gncKVPValue" TYPE "Trail SELECT '");     \
   p = stpcpy (p, sess_str);                                            \
   p = stpcpy (p, "' as sessionGuid, datetime('NOW') as date_changed, " \
                  "'d' as change, objtype as objtype, data as data, type as type, num as
num, denom as denom, ");                   \
   p = stpcpy (p, "* from gncKVPValue" TYPE " WHERE iguid=");           \
   p = stpcpy (p, iguid_str);                                           \
}

Maybe this is a postgres 7.0 bug ??

If you can reproduce the bug, maybe we should look at it more carefully, 
since I can't beleive that its failing to copy the needed data ...

> at the list of object types in event.c and assumed it had to be one of
> those, presumably the type from which the kvp frame came from. 

Nahh, the objtype is sort-of bogus, its there only to allow the SQL
equivalent of an object-oriented 'upcast'.  SQL is only semi-object-oriented, 
and while it allows inheritance, it looses track of the child class during
a downcast. So I store the the type of the child class by hand.

'classtype' might have been a better name.

> But this
> is easy to fix now.

Uhh, will you fix this, or should I ??

--linas


-- 
pub  1024D/01045933 2001-02-01 Linas Vepstas (Labas!) <linas@linas.org>
PGP Key fingerprint = 8305 2521 6000 0B5E 8984  3F54 64A9 9A82 0104 5933