FreeTDS Development Group

["Brian Bruns" <camber AT ais.org>]

> > From: Brian Bruns [mailto:camber AT ais.org]
> > Sent: September 3, 2002 10:24 AM
> > 
> > Jim, I put some verbage in the troubleshooting section about 
> > tsql, but I
> > fear it is already out of date given Freddy's work on domains.  I'll
> > finish up the general documentation for it today.
> 
> Sounds good.  

Submitted a patch to sf with my changes, no access to CVS atm.  

I removed the references to "swap broken dates/money".  1) Figuring out
what server version had what bug was a nightmare 2) it's TDS 4.2 only 3)
"emulate little endian" is the simpler fix.

> > I posted some stuff for compiling the Sybase python module, 
> > i'd like to add that in too.
> 
> Funny, I think I remember seeing that, but it's not in the UG.  I'll check
> my patches directory and mailboxes tonight.  
> 
> > There's the 0x0f cancel packet problem, i think should be fixed before
> > 0.60final. 
> 
> I'm glad to hear you say that.   
> 
> From what I can tell, the "else if (marker == 0)" was added to
> tds_process_cancel to allow for the possibility that a dead socket returns
> no bytes.  Removing that line would resurrect the "loop forever on dead
> socket" bug.  It seems to me that anything expecting data from a socket has
> to distinguish between a packet with no data and a socket with no
> connection.  
> 
> For sure, we don't react to an 0x0f token intelligently; at best we ignore
> it.  What should happen, I think: we should recognize the 0x0f as a cancel
> acknowledgement and record it somewhere as a TDS state, even if we don't use
> the information anywhere yet.  Then, having dealt with the 0x0f packet, we
> keep looking for an 0x04 with the cancel bit set?  

In retrospect, we didn't do a great job of considering this posibility. 
C'est la vie.
 
> > What is the status on conversion?  Any outstanding problems
> > there? (money?).  
> 
> Fixed AFAIK.  Frediano fixed my major complaints on Friday.  Haven't re-run
> my tests myself yet. 
> 
> > Is ODBC working with DBD::ODBC now? There was a 
> > problem as I recall.  Others?
> 
> I don't plan to test ODBC.  

I don't have the time.  Maybe Eric will update us on its viability?
 
> --jkl
>