FreeTDS Development Group

["ZIGLIO, Frediano, VF-IT" <Frediano.Ziglio AT vodafone.com>]

> 
> Thanks Freddy, 
> 
> I'll put this on my TODO.
> 
> I've lost any memory I had of this issue/bug .
> 
> Do we have a test case / program or a more detailed 
> description of the issue
> ?
> 
> merge publication seems to be something to do with replication ?
> 
> Bill
> 
> 

Hi Bill, 
  nice to hear you. I hope you enjoined your holidays..

Well, replication is handled with triggers. Triggers just do a select
and insert/delete on the other end (it capture delete, insert and
update). So you can try using simpler tables with triggers. (cf:
"update/delete returns -1 instead of correct affected rows")

Some sum on others ct-lib issues
1- memory leak patch (cf: "ct_param memory leak")
2- cancellation (cf: "Bugfix for cancelling a query" and "goodread not
so fine")

About point 2. ct_cancel do not handle correctly last parameter (cancel
type). ATTN should just send cancellation, not parse wire. Also
ct_cancel can be called (and are called by sqsh) inside a signal
processing (or think from another thread, it's the same problem) so
ct_results should detect cancellation and return CS_CANCELLED. It's
quite complicated and require some libTDS change. I remember that I
added a TDS_CANCELLED state to libTDS but I didn't write code to use
this state... Problem is that calling ct_cancel "inside" ct_results
(from a signal called inside ct_results) cause state got set to IDLE so
ct_results hangs in goodread. This is also a TODO for ODBC (SQLCancel
can be called from another thread).

freddy77